Bug 1370092 - VERY poor replica write perfomance
Summary: VERY poor replica write perfomance
Keywords:
Status: CLOSED EOL
Alias: None
Product: GlusterFS
Classification: Community
Component: replicate
Version: 3.8
Hardware: x86_64
OS: Linux
medium
urgent
Target Milestone: ---
Assignee: Pranith Kumar K
QA Contact:
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2016-08-25 09:55 UTC by Badalyan Vyacheslav
Modified: 2023-09-14 03:30 UTC (History)
2 users (show)

Fixed In Version:
Doc Type: If docs needed, set a value
Doc Text:
Clone Of:
Environment:
Last Closed: 2017-11-07 10:36:36 UTC
Regression: ---
Mount Type: ---
Documentation: ---
CRM:
Verified Versions:
Embargoed:


Attachments (Terms of Use)

Description Badalyan Vyacheslav 2016-08-25 09:55:38 UTC
Description of problem:
Replica 2 perfomance max is 500MB/s

Version-Release number of selected component (if applicable):
2 equal servers:
- E5-2690 x 2
- 256 GB 1600
- 40GBS infiniband link (Cx-2) p2p link. Connected mode.

How reproducible:
Create 2 RAM disk and create replica 2 volume. mount volume (TCP only, RDMA only). Try to write. Only 400MB/s 

Steps to Reproduce:
on All servers:
- mkdir /mnt/ram
- mount  -t tmpfs -o size=1024M tmpfs /tmp/ram
- mkdir /mnt/ram/b1

gluster vol create ram replica 2 [transport...] s1:/mnt/ram/b1 s2:/mnt/ram/b1
gluster vol start ram

mkdir /tmp/glu
mkdir /tmp/ram
mount -t tmpfs -o size=1024M tmpfs /tmp/ram
mount -t glusterfs [-o transport=rdma] s1:/ram /tmp/glu

dd if=/dev/sda of=/tmp/ram/0.bin bs=128M count=5 #prepare data
dd if=/tmp/ram/0.bin of=/tmp/glu/0.bin bs=128M 

Result - 400-500MB/s
I was try do TCP tunning, use RDMA or TCP only, turn off all perfomance params in vol. I can't get more 500MB/s. Ram disk read/write have 8GB/s
P.S. Many times try read from vol. Read is 500MB/s, 2GB/s, 2GB.

I have BIG and GREAT waits on write on my cluster! In production i use ovirt hypercovered (3 servers, RAID0 8xSSD 1TB, Replica 3 arbiter 1)

Actual results:
Max write speed is 500MB/s

Expected results:
Write speed must be 6-8 GB/s

Comment 1 Badalyan Vyacheslav 2016-08-25 10:01:45 UTC
With write window = 1GB

[root@xintel1 cmdline]# dd if=/tmp/1/1.bin of=/tmp/0/1.bin bs=128M
 скопировано 536870912 байт (537 MB), 0,766988 c, 700 MB/c

[root@xintel1 cmdline]# dd if=/tmp/1/1.bin of=/tmp/0/1.bin bs=128M
 скопировано 536870912 байт (537 MB), 0,807878 c, 665 MB/c

[root@xintel1 cmdline]# dd if=/tmp/1/1.bin of=/tmp/0/1.bin bs=128M
 скопировано 536870912 байт (537 MB), 0,861966 c, 623 MB/c

[root@xintel1 cmdline]# dd if=/tmp/1/1.bin of=/tmp/0/1.bin bs=128M
 скопировано 536870912 байт (537 MB), 0,798036 c, 673 MB/c

[root@xintel1 cmdline]# dd if=/tmp/1/1.bin of=/tmp/0/1.bin bs=128M
 скопировано 536870912 байт (537 MB), 1,59656 c, 336 MB/c

[root@xintel1 cmdline]# dd if=/tmp/1/1.bin of=/tmp/0/1.bin bs=128M
 скопировано 536870912 байт (537 MB), 1,30761 c, 411 MB/c

[root@xintel1 cmdline]# dd if=/tmp/1/1.bin of=/tmp/0/1.bin bs=128M
 скопировано 536870912 байт (537 MB), 1,23974 c, 433 MB/c

[root@xintel1 cmdline]# dd if=/tmp/1/1.bin of=/tmp/0/1.bin bs=512M
 скопировано 536870912 байт (537 MB), 1,89813 c, 283 MB/c

[root@xintel1 cmdline]# dd if=/tmp/1/1.bin of=/tmp/0/1.bin bs=512M
 скопировано 536870912 байт (537 MB), 1,0454 c, 514 MB/c

Comment 2 Badalyan Vyacheslav 2016-08-25 10:26:23 UTC
with nfs mount to gNFS i get 300MB/s

Comment 3 Badalyan Vyacheslav 2016-08-26 09:13:44 UTC
Brick: intel2:/mnt/ram/b1
-------------------------
Cumulative Stats:
   Block Size:             131072b+
 No. of Reads:                   40
No. of Writes:                28672
 %-latency   Avg-latency   Min-Latency   Max-Latency   No. of calls         Fop
 ---------   -----------   -----------   -----------   ------------        ----
      0.00       0.00 us       0.00 us       0.00 us              8     RELEASE
      0.00       0.00 us       0.00 us       0.00 us              6  RELEASEDIR
      0.00       2.83 us       1.00 us       8.00 us              6     OPENDIR
      0.00      19.43 us      14.00 us      25.00 us              7       FLUSH
      0.00      35.50 us      22.00 us      46.00 us              4      STATFS
      0.01      63.86 us      48.00 us     101.00 us              7 REMOVEXATTR
      0.02      45.06 us      11.00 us     129.00 us             16    GETXATTR
      0.03     102.88 us      77.00 us     123.00 us              8        OPEN
      0.05      73.84 us      33.00 us     131.00 us             19       FSTAT
      0.05      26.52 us      13.00 us      91.00 us             58    FINODELK
      0.07     128.56 us     103.00 us     166.00 us             16     XATTROP
      0.11     280.33 us     123.00 us     945.00 us             12     READDIR
      0.16     118.03 us      33.00 us    2967.00 us             40        READ
      0.48     141.82 us      38.00 us     458.00 us            102    FXATTROP
      0.91     555.98 us       6.00 us   21644.00 us             50      LOOKUP
      5.09      18.80 us      11.00 us    1050.00 us           8228     INODELK
     15.93   69172.43 us   65170.00 us   84091.00 us              7    TRUNCATE
     77.09      81.75 us      56.00 us    5169.00 us          28672       WRITE

    Duration: 126 seconds
   Data Read: 5242880 bytes
Data Written: 3758096384 bytes

Interval 1 Stats:
   Block Size:             131072b+
 No. of Reads:                    0
No. of Writes:                20480
 %-latency   Avg-latency   Min-Latency   Max-Latency   No. of calls         Fop
 ---------   -----------   -----------   -----------   ------------        ----
      0.00       0.00 us       0.00 us       0.00 us              5     RELEASE
      0.00      17.60 us      14.00 us      21.00 us              5       FLUSH
      0.01      58.80 us      48.00 us      79.00 us              5 REMOVEXATTR
      0.02      97.80 us      77.00 us     116.00 us              5        OPEN
      0.04      39.10 us      19.00 us      66.00 us             20     INODELK
      0.05      74.71 us      38.00 us     119.00 us             14       FSTAT
      0.05      26.93 us      13.00 us      91.00 us             42    FINODELK
      0.06     124.50 us     103.00 us     162.00 us             10     XATTROP
      0.06     129.80 us      77.00 us     190.00 us             10      LOOKUP
      0.54     148.45 us      38.00 us     458.00 us             76    FXATTROP
     16.84   70477.40 us   65170.00 us   84091.00 us              5    TRUNCATE
     82.32      84.12 us      59.00 us    5169.00 us          20480       WRITE

    Duration: 68 seconds
   Data Read: 0 bytes
Data Written: 2684354560 bytes

Comment 4 Badalyan Vyacheslav 2016-08-26 09:24:45 UTC
In RDMA Only mode:


Brick: intel2:/mnt/ram/b1
-------------------------
Cumulative Stats:
   Block Size:             131072b+
 No. of Reads:                    0
No. of Writes:                32768
 %-latency   Avg-latency   Min-Latency   Max-Latency   No. of calls         Fop
 ---------   -----------   -----------   -----------   ------------        ----
      0.00       0.00 us       0.00 us       0.00 us              8     RELEASE
      0.00       0.00 us       0.00 us       0.00 us              3  RELEASEDIR
      0.00       2.00 us       1.00 us       3.00 us              3     OPENDIR
      0.00      36.00 us      28.00 us      49.00 us              3      STATFS
      0.01      21.12 us      15.00 us      31.00 us              8       FLUSH
      0.01      29.44 us      10.00 us      68.00 us              9    GETXATTR
      0.02      59.75 us      50.00 us      73.00 us              8 REMOVEXATTR
      0.03     112.25 us      83.00 us     134.00 us              8        OPEN
      0.03      29.22 us      14.00 us      64.00 us             32    FINODELK
      0.04      95.69 us      53.00 us     135.00 us             13       FSTAT
      0.05      47.22 us      18.00 us     135.00 us             32     INODELK
      0.05     277.50 us      80.00 us     884.00 us              6     READDIR
      0.08     145.50 us     116.00 us     211.00 us             16     XATTROP
      0.12      90.38 us       7.00 us     200.00 us             39      LOOKUP
      0.28     108.29 us      37.00 us     452.00 us             78    FXATTROP
     17.60   66873.50 us   64404.00 us   70025.00 us              8    TRUNCATE
     81.69      75.79 us      56.00 us    1128.00 us          32768       WRITE

    Duration: 92 seconds
   Data Read: 0 bytes
Data Written: 4294967296 bytes

Interval 2 Stats:
   Block Size:             131072b+
 No. of Reads:                    0
No. of Writes:                16384
 %-latency   Avg-latency   Min-Latency   Max-Latency   No. of calls         Fop
 ---------   -----------   -----------   -----------   ------------        ----
      0.00       0.00 us       0.00 us       0.00 us              4     RELEASE
      0.00      17.25 us      15.00 us      19.00 us              4       FLUSH
      0.01      52.00 us      50.00 us      54.00 us              4 REMOVEXATTR
      0.02      28.75 us      20.00 us      57.00 us             12    FINODELK
      0.03     106.75 us      83.00 us     127.00 us              4        OPEN
      0.04      94.83 us      53.00 us     118.00 us              6       FSTAT
      0.04      37.69 us      18.00 us      67.00 us             16     INODELK
      0.06     118.75 us      67.00 us     173.00 us              8      LOOKUP
      0.07     133.88 us     116.00 us     150.00 us              8     XATTROP
      0.19      92.03 us      37.00 us     405.00 us             32    FXATTROP
     18.00   68204.50 us   66035.00 us   70025.00 us              4    TRUNCATE
     81.53      75.42 us      56.00 us     223.00 us          16384       WRITE

    Duration: 18 seconds
   Data Read: 0 bytes
Data Written: 2147483648 bytes

Comment 5 Badalyan Vyacheslav 2016-08-26 09:41:17 UTC
PERF in RDMA mode:

Samples: 1K of event 'cycles', Event count (approx.): 4212235834
  Children      Self  Command     Shared Object          Symbol                                                                                                                                                                             ◆
+   52,99%     0,07%  glusterfsd  [kernel.kallsyms]      [k] system_call_fastpath                                                                                                                                                           ▒
+   29,23%     0,00%  glusterfsd  libpthread-2.17.so     [.] 0xffff80dd2cab9af3                                                                                                                                                             ▒
+   29,23%     0,00%  glusterfsd  [kernel.kallsyms]      [k] sys_pwrite64                                                                                                                                                                   ▒
+   29,23%     0,08%  glusterfsd  [kernel.kallsyms]      [k] do_sync_write                                                                                                                                                                  ▒
+   29,14%     0,00%  glusterfsd  [kernel.kallsyms]      [k] vfs_write                                                                                                                                                                      ▒
+   29,14%     0,00%  glusterfsd  [kernel.kallsyms]      [k] generic_file_aio_write                                                                                                                                                         ▒
+   28,77%     0,00%  glusterfsd  [kernel.kallsyms]      [k] __generic_file_aio_write                                                                                                                                                       ▒
+   28,58%     0,10%  glusterfsd  [kernel.kallsyms]      [k] generic_file_buffered_write                                                                                                                                                    ▒
+   25,43%     0,00%  glusterfsd  [unknown]              [.] 0000000000000000                                                                                                                                                               ▒
+   18,64%     0,00%  glusterfsd  [kernel.kallsyms]      [k] shmem_write_begin                                                                                                                                                              ▒
+   18,57%     0,62%  glusterfsd  [kernel.kallsyms]      [k] shmem_getpage_gfp                                                                                                                                                              ▒
+   16,32%     0,00%  glusterfsd  [unknown]              [.] 0x0000baadf00d0068                                                                                                                                                             ▒
+   16,32%     0,00%  glusterfsd  [unknown]              [.] 0x00007f22c0009460                                                                                                                                                             ▒
+   16,32%     0,00%  glusterfsd  trash.so               [.] 0xffff80dd39d87063                                                                                                                                                             ▒
+   16,32%     0,00%  glusterfsd  libc-2.17.so           [.] truncate                                                                                                                                                                       ▒
+   16,32%     0,00%  glusterfsd  [kernel.kallsyms]      [k] sys_truncate                                                                                                                                                                   ▒
+   16,32%     0,00%  glusterfsd  [kernel.kallsyms]      [k] do_sys_truncate                                                                                                                                                                ▒
+   16,32%     0,00%  glusterfsd  [kernel.kallsyms]      [k] vfs_truncate                                                                                                                                                                   ▒
+   16,32%     0,00%  glusterfsd  [kernel.kallsyms]      [k] do_truncate                                                                                                                                                                    ▒
+   16,32%     0,00%  glusterfsd  [kernel.kallsyms]      [k] notify_change                                                                                                                                                                  ▒
+   16,32%     0,00%  glusterfsd  [kernel.kallsyms]      [k] shmem_setattr                                                                                                                                                                  ▒
+   16,32%     0,00%  glusterfsd  [kernel.kallsyms]      [k] shmem_truncate_range                                                                                                                                                           ▒
+   16,32%     0,23%  glusterfsd  [kernel.kallsyms]      [k] shmem_undo_range                                                                                                                                                               ▒
+    7,84%     7,76%  glusterfsd  libpthread-2.17.so     [.] pthread_spin_lock                                                                                                                                                              ▒
+    7,71%     0,17%  glusterfsd  [kernel.kallsyms]      [k] shmem_alloc_page                                                                                                                                                               ▒
+    7,45%     0,00%  glusterfsd  [kernel.kallsyms]      [k] alloc_pages_vma                                                                                                                                                                ▒
+    7,39%     0,30%  glusterfsd  [kernel.kallsyms]      [k] truncate_inode_page                                                                                                                                                            ▒
+    7,34%     0,88%  glusterfsd  [kernel.kallsyms]      [k] release_pages                                                                                                                                                                  ▒
+    7,19%     0,00%  glusterfsd  [kernel.kallsyms]      [k] __pagevec_release                                                                                                                                                              ▒
+    6,89%     0,68%  glusterfsd  [kernel.kallsyms]      [k] __alloc_pages_nodemask                                                                                                                                                         ▒
+    6,70%     0,16%  glusterfsd  [kernel.kallsyms]      [k] delete_from_page_cache                                                                                                                                                         ▒
+    6,42%     6,42%  glusterfsd  [kernel.kallsyms]      [k] copy_user_generic_string                                                                                                                                                       ▒
+    5,84%     0,16%  glusterfsd  [kernel.kallsyms]      [k] free_hot_cold_page_list                                                                                                                                                        ▒
+    5,71%     1,95%  glusterfsd  [kernel.kallsyms]      [k] get_page_from_freelist                                                                                                                                                         ▒
+    5,61%     0,28%  glusterfsd  [kernel.kallsyms]      [k] free_hot_cold_page                                                                                                                                                             ▒
+    4,74%     0,00%  glusterfsd  [kernel.kallsyms]      [k] mem_cgroup_uncharge_cache_page                                                                                                                                                 ▒
+    4,67%     4,31%  glusterfsd  [kernel.kallsyms]      [k] __mem_cgroup_uncharge_common                                                                                                                                                   ▒
+    4,60%     0,00%  glusterfsd  libmlx4-rdmav2.so      [.] 0xffff80dd426db780                                                                                                                                                             ▒
+    4,59%     0,07%  glusterfsd  [kernel.kallsyms]      [k] mem_cgroup_cache_charge                                                                                                                                                        ▒
+    4,46%     0,00%  glusterfsd  [kernel.kallsyms]      [k] mem_cgroup_charge_common                                                                                                                                                       ▒
+    4,38%     3,21%  glusterfsd  [kernel.kallsyms]      [k] free_pcppages_bulk                                                                                                                                                             ▒
+    4,32%     3,81%  glusterfsd  [kernel.kallsyms]      [k] __mem_cgroup_commit_charge                                                                                                                                                     ▒
+    4,20%     0,00%  glusterfsd  [unknown]              [.] 0x00007f22b0003d50                                                                                                                                                             ▒
+    3,59%     3,59%  glusterfsd  [kernel.kallsyms]      [k] __list_del_entry                                                                                                                                                               ▒
+    3,56%     0,06%  glusterfsd  [kernel.kallsyms]      [k] list_del                                                                                                                                                                       ▒
+    3,35%     0,00%  glusterfsd  libpthread-2.17.so     [.] 0xffff80dd2cab922d              


Perf in TCP Mode:

Samples: 9K of event 'cycles', Event count (approx.): 29918011566
  Children      Self  Command     Shared Object          Symbol
+   60,45%     0,04%  glusterfsd  [kernel.kallsyms]      [k] system_call_fastpath
+   25,89%     0,00%  glusterfsd  [unknown]              [.] 0000000000000000
+   22,86%     0,10%  glusterfsd  [kernel.kallsyms]      [k] do_readv_writev
+   22,33%     0,07%  glusterfsd  [kernel.kallsyms]      [k] do_sync_readv_writev
+   21,95%     0,00%  glusterfsd  libpthread-2.17.so     [.] 0xffff80a668a10af3
+   21,92%     0,04%  glusterfsd  [kernel.kallsyms]      [k] sys_pwrite64
+   21,85%     0,08%  glusterfsd  [kernel.kallsyms]      [k] vfs_write
+   21,47%     0,03%  glusterfsd  [kernel.kallsyms]      [k] do_sync_write
+   21,40%     0,04%  glusterfsd  [kernel.kallsyms]      [k] generic_file_aio_write
+   21,35%     0,03%  glusterfsd  [kernel.kallsyms]      [k] __generic_file_aio_write
+   21,00%     0,13%  glusterfsd  [kernel.kallsyms]      [k] generic_file_buffered_write
+   15,55%     0,14%  glusterfsd  libc-2.17.so           [.] __libc_readv
+   15,17%     0,07%  glusterfsd  [kernel.kallsyms]      [k] sys_readv
+   15,07%     0,00%  glusterfsd  [kernel.kallsyms]      [k] vfs_readv
+   14,65%     0,03%  glusterfsd  [kernel.kallsyms]      [k] sock_aio_read
+   14,59%     0,04%  glusterfsd  [kernel.kallsyms]      [k] sock_aio_read.part.7
+   14,48%     0,02%  glusterfsd  [kernel.kallsyms]      [k] inet_recvmsg
+   14,44%     0,24%  glusterfsd  [kernel.kallsyms]      [k] tcp_recvmsg
+   11,92%     0,11%  glusterfsd  [kernel.kallsyms]      [k] shmem_write_begin
+   11,85%    11,02%  glusterfsd  libpthread-2.17.so     [.] pthread_spin_lock
+   11,70%     0,74%  glusterfsd  [kernel.kallsyms]      [k] shmem_getpage_gfp
+   11,50%    11,50%  glusterfsd  [kernel.kallsyms]      [k] copy_user_generic_string
+   10,85%     0,00%  glusterfsd  [unknown]              [.] 0x0000000000000004
+    9,88%     0,16%  glusterfsd  [kernel.kallsyms]      [k] tcp_transmit_skb
+    9,49%     0,04%  glusterfsd  [kernel.kallsyms]      [k] ip_queue_xmit
+    9,34%     0,00%  glusterfsd  [kernel.kallsyms]      [k] ip_local_out_sk
+    8,37%     0,00%  glusterfsd  [unknown]              [.] 0x000000000001fcbc
+    8,14%     0,03%  glusterfsd  libc-2.17.so           [.] __libc_writev
+    7,93%     0,08%  glusterfsd  [kernel.kallsyms]      [k] sys_writev
+    7,83%     0,00%  glusterfsd  [kernel.kallsyms]      [k] vfs_writev
+    7,59%     0,05%  glusterfsd  [kernel.kallsyms]      [k] sock_aio_write
+    7,54%     0,01%  glusterfsd  [kernel.kallsyms]      [k] inet_sendmsg
+    7,45%     0,04%  glusterfsd  [kernel.kallsyms]      [k] intel_map_page
+    7,40%     0,16%  glusterfsd  [kernel.kallsyms]      [k] tcp_sendmsg
+    7,31%     0,06%  glusterfsd  [kernel.kallsyms]      [k] __intel_map_single
+    7,26%     0,04%  glusterfsd  [kernel.kallsyms]      [k] ip_output

event_analyzing_sample:

In trace_end:                                                                                                                                                                                                                               ▒
                                                                                                                                                                                                                                            ▒
There is 13548 records in gen_events table                                                                                                                                                                                                  ▒
Statistics about the general events grouped by thread/symbol/dso:                                                                                                                                                                           ▒
                                                                                                                                                                                                                                            ▒
                                                                                                                                                                                                                                            ▒
            comm   number        histogram                                                                                                                                                                                                  ▒
==========================================                                                                                                                                                                                                  ▒
      glusterfsd    13548     ##############                                                                                                                                                                                                ▒
                                                                                                                                                                                                                                            ▒
                          symbol   number        histogram                                                                                                                                                                                  ▒
==========================================================                                                                                                                                                                                  ▒
           native_write_msr_safe     1764     ###########                                                                                                                                                                                   ▒
               pthread_spin_lock     1211     ###########                                                                                                                                                                                   ▒
        copy_user_generic_string     1173     ###########                                                                                                                                                                                   ▒
                  Unknown_symbol      658     ##########                                                                                                                                                                                    ▒
                __list_del_entry      316     #########                                                                                                                                                                                     ▒
          _raw_spin_lock_irqsave      277     #########                                                                                                                                                                                     ▒
                         _gf_msg      275     #########                                                                                                                                                                                     ▒
                         rb_prev      270     #########                                                                                                                                                                                     ▒
    __mem_cgroup_uncharge_common      267     #########                                                                                                                                                                                     ▒
              free_pcppages_bulk      230     ########                                                                                                                                                                                      ▒
      __mem_cgroup_commit_charge      225     ########                                                                                                                                                                                      ▒
              mark_page_accessed      165     ########                                                                                                                                                                                      ▒
                        put_page      125     #######                                                                                                                                                                                       ▒
       get_pageblock_flags_group      124     #######                                                                                                                                                                                       ▒
                        vfprintf      122     #######                                                                                                                                                                                       ▒
             __radix_tree_lookup      119     #######                                                                                                                                                                                       ▒
                   __memset_sse2      114     #######                                                                                                                                                                                       ▒
       __glusterfs_this_location      110     #######                                                                                                                                                                                       ▒
                     _int_malloc       98     #######                                                                                                                                                                                       ▒
                  _raw_spin_lock       91     #######                                                                                                                                                                                       ▒
          get_page_from_freelist       87     #######                                                                                                                                                                                       ▒
                   release_pages       77     #######                                                                                                                                                                                       ▒
                         do_csum       74     #######                                                                                                                                                                                       ▒
                __find_get_pages       71     #######                                                                                                                                                                                       ▒
             pthread_getspecific       64     #######                                                                                                                                                                                       ▒
               shmem_getpage_gfp       63     ######                                                                                                                                                                                        ▒
              lookup_page_cgroup       60     ######                                                                                                                                                                                        ▒
              pthread_mutex_lock       59     ######                                                                                                                                                                                        ▒
                       _int_free       55     ######                                                                                                                                                                                        ▒
                      alloc_iova       54     ######                                                                                                                                                                                        ▒
                         mem_get       54     ######                                                                                                                                                                                        ▒
                      __list_add       53     ######                                                                                                                                                                                        ▒
              finish_task_switch       50     ######                                                                                                                                                                                        ▒
          __alloc_pages_nodemask       49     ######                                                                                                                                                                                        ▒
             __memcpy_ssse3_back       49     ######                                                                                                                                                                                        ▒
            gf_mem_set_acct_info       49     ######                                                                                                                                                                                        ▒
              _raw_spin_lock_irq       47     ######                                                                                                                                                                                        ▒
              free_hot_cold_page       47     ######                                                                                                                                                                                        ▒
                       __gf_free       45     ######                                                                                                                                                                                        ▒
           __mod_zone_page_state       45     ######                                                                                                                                                                                        ▒
            pthread_mutex_unlock       44     ######                                                                                                                                                                                        ▒
                  page_waitqueue       43     ######                                                                                                                                                                                        ▒
                   __wake_up_bit       41     ######                                                                                                                                                                                        ▒
                   __libc_calloc       40     ######                                                                                                                                                                                        ▒
                           kfree       40     ######                                                                                                                                                                                        ▒
                __inc_zone_state       39     ######                                                                                                                                                                                        ▒
                    ipt_do_table       38     ######

Comment 6 Badalyan Vyacheslav 2016-08-26 09:55:02 UTC
[root@xintel2 ~]# perf stat -a -g  -p 110232
^C
 Performance counter stats for process id '110232':

      33604,217420      task-clock (msec)         #    3,467 CPUs utilized            (1,90%)
           444 787      context-switches          #    0,013 M/sec                    (1,90%)
            39 759      cpu-migrations            #    0,001 M/sec                    (1,90%)
                 0      page-faults               #    0,000 K/sec                    (1,90%)
    12 270 079 483      cycles                    #    0,365 GHz                      (1,90%)
     6 796 842 214      stalled-cycles-frontend   #   55,39% frontend cycles idle     (0,01%)
     5 270 033 500      stalled-cycles-backend    #   42,95% backend  cycles idle     (0,01%)
     2 911 092 877      instructions              #    0,24  insns per cycle
                                                  #    2,33  stalled cycles per insn  (0,01%)
       394 494 858      branches                  #   11,739 M/sec                    (0,00%)
     <not counted>      branch-misses              (0,00%)

       9,692958135 seconds time elapsed

Comment 7 Badalyan Vyacheslav 2016-08-26 10:13:06 UTC
Detalid TCP mode 1 file DD test on node2:

-   62,39%     0,01%  glusterfsd  [k] system_call_fastpath                                                                                                                                                                                  ▒
   - system_call_fastpath                                                                                                                                                                                                                   ▒
        35,63% 0xeaf3                                                                                                                                                                                                                       ▒
      + 26,82% __libc_readv                                                                                                                                                                                                                 ▒
      + 12,85% __libc_writev                                                                                                                                                                                                                ◆
      + 11,56% truncate                                                                                                                                                                                                                     ▒
      + 3,74% 0xf72c3                                                                                                                                                                                                                       ▒
      + 2,58% pthread_cond_timedwait@@GLIBC_2.3.2                                                                                                                                                                                           ▒
      + 2,32% fgetxattr                                                                                                                                                                                                                     ▒
      + 1,07% __lll_lock_wait                                                                                                                                                                                                               ▒
      + 1,07% pthread_cond_signal@@GLIBC_2.3.2                                                                                                                                                                                              ▒
      + 0,86% __lll_unlock_wake                                                                                                                                                                                                             ▒
      + 0,76% __fxstat64                                                                                                                                                                                                                    ▒
      + 0,56% epoll_ctl                                                                                                                                                                                                                     ▒
-   24,22%     0,08%  glusterfsd  [k] do_readv_writev                                                                                                                                                                                       ▒
   - do_readv_writev                                                                                                                                                                                                                        ▒
      + 67,43% vfs_readv                                                                                                                                                                                                                    ▒
      + 32,54% vfs_writev                                                                                                                                                                                                                   ▒
-   23,75%     0,14%  glusterfsd  [k] do_sync_readv_writev                                                                                                                                                                                  ▒
   - do_sync_readv_writev                                                                                                                                                                                                                   ▒
      - do_readv_writev                                                                                                                                                                                                                     ▒
         + 67,54% vfs_readv                                                                                                                                                                                                                 ▒
         + 32,46% vfs_writev                                                                                                                                                                                                                ▒
-   22,21%     0,00%  glusterfsd  [k] sys_pwrite64                                                                                                                                                                                          ▒
     sys_pwrite64                                                                                                                                                                                                                           ▒
     system_call_fastpath                                                                                                                                                                                                                   ▒
     0xeaf3                                                                                                                                                                                                                                 ▒
-   22,17%     0,03%  glusterfsd  [k] vfs_write                                                                                                                                                                                             ▒
     vfs_write                                                                                                                                                                                                                              ▒
     sys_pwrite64                                                                                                                                                                                                                           ▒
     system_call_fastpath                                                                                                                                                                                                                   ▒
     0xeaf3                                                                                                                                                                                                                                 ▒
-   21,99%     0,03%  glusterfsd  [k] do_sync_write                                                                                                                                                                                         ▒
   - do_sync_write                                                                                                                                                                                                                          ▒
      + vfs_write                                                                                                                                                                                                                           ▒
-   21,89%     0,05%  glusterfsd  [k] generic_file_aio_write                                                                                                                                                                                ▒
     generic_file_aio_write                                                                                                                                                                                                                 ▒
     do_sync_write                                                                                                                                                                                                                          ▒
     vfs_write                                                                                                                                                                                                                              ▒
     sys_pwrite64                                                                                                                                                                                                                           ▒
     system_call_fastpath                                                                                                                                                                                                                   ▒
     0xeaf3                                                                                                                                                                                                                                 ▒
-   21,79%     0,00%  glusterfsd  [k] __generic_file_aio_write                                                                                                                                                                              ▒
     __generic_file_aio_write                                                                                                                                                                                                               ▒
     generic_file_aio_write                                                                                                                                                                                                                 ▒
     do_sync_write                                                                                                                                                                                                                          ▒
     vfs_write                                                                                                                                                                                                                              ▒
     sys_pwrite64                                                                                                                                                                                                                           ▒
     system_call_fastpath                                                                                                                                                                                                                   ▒
     0xeaf3                                                                                                                                                                                                                                 ▒
-   21,40%     0,16%  glusterfsd  [k] generic_file_buffered_write                                                                                                                                                                           ▒
     generic_file_buffered_write                                                                                                                                                                                                            ▒
     __generic_file_aio_write                                                                                                                                                                                                               ▒
     generic_file_aio_write                                                                                                                                                                                                                 ▒
     do_sync_write                                                                                                                                                                                                                          ▒
     vfs_write                                                                                                                                                                                                                              ▒
     sys_pwrite64                                                                                                                                                                                                                           ▒
     system_call_fastpath                                                                                                                                                                                                                   ▒
     0xeaf3                                                                                                                                                                                                                                 ▒
+   16,73%     0,05%  glusterfsd  [k] sys_readv

Comment 8 Badalyan Vyacheslav 2016-08-26 11:00:11 UTC
with linux AIO enabled..

-   64,12%     0,04%  glusterfsd  [kernel.vmlinux]       [k] system_call_fastpath                                                                                                                                                           ▒
   - system_call_fastpath                                                                                                                                                                                                                   ▒
      + 44,99% io_submit                                                                                                                                                                                                                    ▒
      + 27,21% __libc_readv                                                                                                                                                                                                                 ▒
      + 10,80% __libc_writev                                                                                                                                                                                                                ▒
      + 4,54% 0xf72c3                                                                                                                                                                                                                       ▒
      + 3,16% pthread_cond_timedwait@@GLIBC_2.3.2                                                                                                                                                                                           ▒
      + 2,04% fgetxattr                                                                                                                                                                                                                     ▒
      + 1,90% 0x644                                                                                                                                                                                                                         ▒
      + 1,46% pthread_cond_signal@@GLIBC_2.3.2                                                                                                                                                                                              ▒
      + 1,15% __lll_lock_wait                                                                                                                                                                                                               ▒
      + 0,94% epoll_ctl                                                                                                                                                                                                                     ▒
      + 0,82% __lll_unlock_wake                                                                                                                                                                                                             ▒
      + 0,80% __fxstat64                                                                                                                                                                                                                    ▒
+   48,58%     0,00%  glusterfsd  [unknown]              [.] 0000000000000000                                                                                                                                                               ▒
+   29,09%     0,04%  glusterfsd  libaio.so.1.0.1        [.] io_submit                                                                                                                                                                      ▒
+   28,85%     0,01%  glusterfsd  [kernel.vmlinux]       [k] sys_io_submit                                                                                                                                                                  ▒
+   28,60%     0,18%  glusterfsd  [kernel.vmlinux]       [k] do_io_submit                                                                                                                                                                   ▒
+   27,14%     0,02%  glusterfsd  [kernel.vmlinux]       [k] generic_file_aio_write                                                                                                                                                         ▒
+   27,05%     0,10%  glusterfsd  [kernel.vmlinux]       [k] __generic_file_aio_write                                                                                                                                                       ▒
+   26,47%     0,10%  glusterfsd  [kernel.vmlinux]       [k] generic_file_buffered_write                                                                                                                                                    ▒
+   24,14%     0,12%  glusterfsd  [kernel.vmlinux]       [k] do_readv_writev                                                                                                                                                                ▒
+   23,30%     0,13%  glusterfsd  [kernel.vmlinux]       [k] do_sync_readv_writev                                                                                                                                                           ▒
+   17,73%     0,03%  glusterfsd  libc-2.17.so           [.] __libc_readv                                                                                                                                                                   ▒
+   17,45%     0,13%  glusterfsd  [kernel.vmlinux]       [k] sys_readv                                                                                                                                                                      ▒
+   17,26%     0,00%  glusterfsd  [kernel.vmlinux]       [k] vfs_readv                                                                                                                                                                      ▒
+   16,55%     0,01%  glusterfsd  [kernel.vmlinux]       [k] sock_aio_read                                                                                                                                                                  ▒
+   16,50%     0,02%  glusterfsd  [kernel.vmlinux]       [k] sock_aio_read.part.7                                                                                                                                                           ▒
+   16,49%     0,02%  glusterfsd  [kernel.vmlinux]       [k] inet_recvmsg                                                                                                                                                                   ▒
+   16,36%     0,20%  glusterfsd  [kernel.vmlinux]       [k] tcp_recvmsg                                                                                                                                                                    ▒
+   14,58%     0,03%  glusterfsd  [kernel.vmlinux]       [k] shmem_write_begin                                                                                                                                                              ▒
+   14,23%     0,74%  glusterfsd  [kernel.vmlinux]       [k] shmem_getpage_gfp                                                                                                                                                              ▒
+   13,29%    13,29%  glusterfsd  [kernel.vmlinux]       [k] copy_user_generic_string                                                                                                                                                       ▒
+    9,79%     0,00%  glusterfsd  [unknown]              [.] 0x0000000000000004                                                                                                                                                             ▒
+    9,60%     0,00%  glusterfsd  [unknown]              [.] 0x000000000001fcbc                                                                                                                                                             ▒
+    9,19%     0,24%  glusterfsd  [kernel.vmlinux]       [k] intel_map_page                                                                                                                                                                 ▒
+    9,16%     0,19%  glusterfsd  [kernel.vmlinux]       [k] tcp_transmit_skb                                                                                                                                                               ▒
+    8,89%     0,11%  glusterfsd  [kernel.vmlinux]       [k] __intel_map_single                                                                                                                                                             ▒
+    8,65%     0,04%  glusterfsd  [kernel.vmlinux]       [k] ip_queue_xmit                                                                                                                                                                  ▒
+    8,56%     0,00%  glusterfsd  [kernel.vmlinux]       [k] ip_local_out_sk                                                                                                                                                                ▒
+    8,10%     0,00%  glusterfsd  [kernel.vmlinux]       [k] intel_alloc_iova                                                                                                                                                               ▒
+    8,02%     0,03%  glusterfsd  [kernel.vmlinux]       [k] common_interrupt                                                                                                                                                               ▒
+    8,01%     0,00%  glusterfsd  [kernel.vmlinux]       [k] do_softirq                                                                                                                                                                     ▒
+    8,01%     0,00%  glusterfsd  [kernel.vmlinux]       [k] call_softirq                                                                                                                                                                   ▒
+    8,01%     0,01%  glusterfsd  [kernel.vmlinux]       [k] __do_softirq                                                                                                                                                                   ▒
+    7,97%     0,00%  glusterfsd  [kernel.vmlinux]       [k] do_IRQ                                                                                                                                                                         ▒
+    7,85%     0,00%  glusterfsd  [kernel.vmlinux]       [k] irq_exit                                                                                                                                                                       ▒
+    7,82%     0,00%  glusterfsd  [kernel.vmlinux]       [k] net_rx_action                                                                                                                                                                  ▒
+    7,70%     0,00%  glusterfsd  [ib_ipoib]             [k] ipoib_poll                                                                                                                                                                     ▒
+    7,66%     0,65%  glusterfsd  [kernel.vmlinux]       [k] alloc_iova                                                                                                                                                                     ▒
+    7,46%     0,11%  glusterfsd  [ib_ipoib]             [k] ipoib_cm_handle_rx_wc                                                                                                                                                          ▒
+    7,31%     6,60%  glusterfsd  libpthread-2.17.so     [.] pthread_spin_lock                                                                                                                                                              ▒
+    7,15%     0,00%  glusterfsd  [unknown]              [.] 0x00007fa734000c70                                                                                                                                                             ▒
+    7,14%     0,04%  glusterfsd  [kernel.vmlinux]       [k] ip_output

Comment 9 Badalyan Vyacheslav 2016-08-26 11:17:52 UTC
Where good perf counters.... where is bottleneck?


[root@xintel1 ~]# gluster volume top ram write-perf bs 104857600 count 2 list-cnt 0
Brick: intel1:/mnt/ram/b1
Throughput 1754.70 MBps time 0.1195 secs
Brick: intel2:/mnt/ram/b1
Throughput 1989.71 MBps time 0.1054 secs
[root@xintel1 ~]# gluster volume top ram write-perf bs 104857600 count 2 list-cnt 0
Brick: intel1:/mnt/ram/b1
Throughput 1705.35 MBps time 0.1230 secs
Brick: intel2:/mnt/ram/b1
Throughput 1993.51 MBps time 0.1052 secs


[root@xintel1 ~]# gluster volume top ram write-perf
Brick: intel1:/mnt/ram/b1
MBps Filename                                        Time
==== ========                                        ====
2383 /1.bin                                          2016-08-26 11:03:05.619903
2299 /1.bin                                          2016-08-26 10:59:47.175730
2259 /1.bin                                          2016-08-26 10:56:24.305102
2114 /2.bin                                          2016-08-26 11:11:26.216565
1795 /2.bin                                          2016-08-26 11:10:54.378487
Brick: intel2:/mnt/ram/b1
MBps Filename                                        Time
==== ========                                        ====
2570 /1.bin                                          2016-08-26 11:02:56.438934
2221 /2.bin                                          2016-08-26 11:11:30.51475
2184 /1.bin                                          2016-08-26 10:57:59.271729
2184 /1.bin                                          2016-08-26 10:55:42.932653
2080 /2.bin                                          2016-08-26 11:10:55.551528

Comment 10 Niels de Vos 2016-09-12 05:37:43 UTC
All 3.8.x bugs are now reported against version 3.8 (without .x). For more information, see http://www.gluster.org/pipermail/gluster-devel/2016-September/050859.html

Comment 11 Pranith Kumar K 2016-10-14 11:03:39 UTC
What is the performance you see on normal NFS mount without gluster involved? Could you do that test and let us no?

Comment 12 Niels de Vos 2017-11-07 10:36:36 UTC
This bug is getting closed because the 3.8 version is marked End-Of-Life. There will be no further updates to this version. Please open a new bug against a version that still receives bugfixes if you are still facing this issue in a more current release.

Comment 13 Red Hat Bugzilla 2023-09-14 03:30:06 UTC
The needinfo request[s] on this closed bug have been removed as they have been unresolved for 1000 days


Note You need to log in before you can comment on or make changes to this bug.