Bug 2365154 - Input/Output error using fio against CephFS backed NFS mount [NEEDINFO]
Summary: Input/Output error using fio against CephFS backed NFS mount
Keywords:
Status: CLOSED ERRATA
Alias: None
Product: Red Hat Ceph Storage
Classification: Red Hat Storage
Component: NFS-Ganesha
Version: 8.0
Hardware: x86_64
OS: Linux
unspecified
urgent
Target Milestone: ---
: 8.1
Assignee: Frank Filz
QA Contact: Manisha Saini
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2025-05-08 20:13 UTC by Eric Smith
Modified: 2025-06-28 03:20 UTC (History)
18 users (show)

Fixed In Version: nfs-ganesha-6.5-15.el9cp; rhceph-container-8-468
Doc Type: If docs needed, set a value
Doc Text:
Clone Of:
Environment:
Last Closed: 2025-06-26 12:31:40 UTC
Embargoed:
pdhange: needinfo? (cattelan)
pdhange: needinfo? (cattelan)


Attachments (Terms of Use)
Ganesha logs with FSAL = DEBUG_FULL turned on. (1.52 MB, application/octet-stream)
2025-05-13 00:22 UTC, Russell Cattelan
no flags Details
Latest debug run with EVENTS enabled (3.50 MB, text/plain)
2025-05-14 19:31 UTC, Russell Cattelan
no flags Details
capture file, fio results, and notes showing NFS4_OK write reply with count=0 (21.31 KB, application/gzip)
2025-05-15 14:31 UTC, jeff.a.smith
no flags Details
Logs from ganesha debug build 2 (10.29 MB, application/gzip)
2025-05-16 23:58 UTC, Russell Cattelan
no flags Details
ganesha logs from the debug image without zerocopy and FSAL async features (77.81 KB, text/plain)
2025-05-22 22:15 UTC, jeff.a.smith
no flags Details
Latest run with async turned off. (7.34 MB, application/gzip)
2025-05-23 20:37 UTC, Russell Cattelan
no flags Details


Links
System ID Private Priority Status Summary Last Updated
Red Hat Issue Tracker RHCEPH-11366 0 None None None 2025-05-08 20:14:55 UTC
Red Hat Product Errata RHSA-2025:9775 0 None None None 2025-06-26 12:31:45 UTC

Description Eric Smith 2025-05-08 20:13:59 UTC
Description of problem:
Getting error message fio: io_u error on file /mnt/share-test-0/csi-helper-yy346mdx: Input/output error: write offset=525594624, buflen=4096 on mounted share for fio runs.

Version-Release number of selected component (if applicable):
ceph version 19.2.0-124.1.hotfix.bz2357486.el9cp (1d6c39dcc35466271ef633ccfd91e51cc792656b) squid (stable)

How reproducible:
100%

Steps to Reproduce:
1. Create a Ganesha NFS mount
2. Mount the export
3. Run the following FIO:

echo $bs ; echo $njobs ; echo $iodepth
10K
16
128

cat include.fio
[global]
#direct=1
rw=randwrite

[test0]
direct=1
rwmixread=50
bs=${bs}
iodepth=${iodepth}
numjobs=${njobs}
ioengine=libaio
lat_percentiles=1
unified_rw_reporting=1
group_reporting=1
size=1G

while true; do fio include.fio --filename=$(pwd)/test0; done

Actual results:
io_u error on file /mnt/share-test-0/csi-helper-yy346mdx: Input/output error: write offset=525594624, buflen=4096 on mounted share for fio runs.

Expected results:
Successful I/O test

Additional info:

Comment 1 Eric Smith 2025-05-08 20:16:25 UTC
Error in kern.log
NFS: Server wrote zero bytes, expected 4096.

Comment 2 Manisha Saini 2025-05-11 02:06:56 UTC
Hi Eric,

Can you please share the ganesha logs to check on failure. Also, can you also share "ceph -s" output. 
Do we have sufficient space on disk to complete the fio profile?

I ran the same test, but the fio completed without any issues

Could you please share the Ganesha logs to investigate the failure further? 
Can you confirm whether there's enough available space on the disk to complete the FIO workload? It would be helpful to see the output of ceph -s to look for any osd related issues.

For reference, I ran the same FIO profile on my setup and it completed successfully without any issues.

# rpm -qa | grep nfs
libnfsidmap-2.5.4-27.el9_5.1.x86_64
nfs-utils-2.5.4-27.el9_5.1.x86_64
nfs-ganesha-selinux-6.5-13.el9cp.noarch
nfs-ganesha-6.5-13.el9cp.x86_64
nfs-ganesha-ceph-6.5-13.el9cp.x86_64
nfs-ganesha-rados-grace-6.5-13.el9cp.x86_64
nfs-ganesha-rados-urls-6.5-13.el9cp.x86_64
nfs-ganesha-rgw-6.5-13.el9cp.x86_64

# ceph --version
ceph version 19.2.1-188.el9cp (834ac46f780fbdc2ac4ba4851a36db6df3c1aa6f) squid (stable)

[ceph: root@cali013 /]# ceph nfs export ls nfsganesha
[
  "/ganesha1"
]

[ceph: root@cali013 /]# ceph nfs export info nfsganesha /ganesha1
{
  "access_type": "RW",
  "clients": [],
  "cluster_id": "nfsganesha",
  "export_id": 1,
  "fsal": {
    "cmount_path": "/",
    "fs_name": "cephfs",
    "name": "CEPH",
    "user_id": "nfs.nfsganesha.cephfs.2c1043d4"
  },
  "path": "/volumes/ganeshagroup/ganesha1/1ffcbfcd-a03e-4311-a39c-2d2d912cb273",
  "protocols": [
    3,
    4
  ],
  "pseudo": "/ganesha1",
  "security_label": true,
  "squash": "none",
  "transports": [
    "TCP"
  ]
}


Client
-----

# cat include.fio
[global]
rw=randwrite
ioengine=libaio
lat_percentiles=1
unified_rw_reporting=1
group_reporting=1
size=1G

[test0]
direct=1
rwmixread=50
bs=4k        # Default, can be overridden
iodepth=1    # Default, can be overridden
numjobs=1    # Default, can be overridden


# fio include.fio --section=test0 --bs=10k --iodepth=128 --numjobs=16 --filename=/mnt/ganesha/fio_file
test0: (g=0): rw=randwrite, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=libaio, iodepth=1
fio-3.35
Starting 1 process
test0: Laying out IO file (1 file / 1024MiB)
Jobs: 1 (f=1): [w(1)][100.0%][r=2892KiB/s][r=723 IOPS][eta 00m:00s]
test0: (groupid=0, jobs=1): err= 0: pid=8732: Sun May 11 01:54:42 2025
  mixed: IOPS=714, BW=2859KiB/s (2928kB/s)(1024MiB/366759msec)
    slat (usec): min=4, max=6680, avg=36.53, stdev=21.65
    clat (usec): min=432, max=10780, avg=1358.42, stdev=170.85
     lat (usec): min=960, max=10833, avg=1394.95, stdev=172.80
    clat percentiles (usec):
     |  1.00th=[ 1123],  5.00th=[ 1172], 10.00th=[ 1205], 20.00th=[ 1237],
     | 30.00th=[ 1270], 40.00th=[ 1303], 50.00th=[ 1336], 60.00th=[ 1369],
     | 70.00th=[ 1401], 80.00th=[ 1450], 90.00th=[ 1532], 95.00th=[ 1614],
     | 99.00th=[ 1909], 99.50th=[ 2073], 99.90th=[ 2671], 99.95th=[ 3032],
     | 99.99th=[ 5997]
    lat percentiles (usec):
     |  1.00th=[ 1156],  5.00th=[ 1205], 10.00th=[ 1237], 20.00th=[ 1287],
     | 30.00th=[ 1303], 40.00th=[ 1336], 50.00th=[ 1369], 60.00th=[ 1401],
     | 70.00th=[ 1434], 80.00th=[ 1483], 90.00th=[ 1565], 95.00th=[ 1663],
     | 99.00th=[ 1942], 99.50th=[ 2114], 99.90th=[ 2737], 99.95th=[ 3064],
     | 99.99th=[ 6128]
   bw (  KiB/s): min= 2368, max= 3160, per=100.00%, avg=2860.38, stdev=100.58, samples=733
   iops        : min=  592, max=  790, avg=715.08, stdev=25.15, samples=733
  lat (usec)   : 500=0.01%, 750=0.01%, 1000=0.01%
  lat (msec)   : 2=99.32%, 4=0.66%, 10=0.02%, 20=0.01%
  cpu          : usr=0.63%, sys=1.14%, ctx=505001, majf=0, minf=10
  IO depths    : 1=100.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%, >=64=0.0%
     submit    : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
     complete  : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
     issued rwts: total=262144,0,0,0 short=0,0,0,0 dropped=0,0,0,0
     latency   : target=0, window=0, percentile=100.00%, depth=1

Run status group 0 (all jobs):
  MIXED: bw=2859KiB/s (2928kB/s), 2859KiB/s-2859KiB/s (2928kB/s-2928kB/s), io=1024MiB (1074MB), run=366759-366759msec


[root@ceph-nfsclusterlive-bs2b7t-node6 home]# df -hT
Filesystem            Type      Size  Used Avail Use% Mounted on
devtmpfs              devtmpfs  4.0M     0  4.0M   0% /dev
tmpfs                 tmpfs     1.8G     0  1.8G   0% /dev/shm
tmpfs                 tmpfs     732M   28M  704M   4% /run
/dev/vda4             xfs        39G  2.5G   37G   7% /
/dev/vda3             xfs       960M  176M  785M  19% /boot
/dev/vda2             vfat      200M  7.1M  193M   4% /boot/efi
tmpfs                 tmpfs     366M     0  366M   0% /run/user/0
10.8.130.16:/ganesha1 nfs4       22T  1.0G   22T   1% /mnt/ganesha

Comment 3 jeff.a.smith 2025-05-12 16:00:50 UTC
Based on the NFS client-side log message emitted, we know that the EIO is generated by the client, triggered by the NFS server sending a write reply with no error on the wire, but 0 bytes written.   The linux client thinks that is server misbehavior and returns EIO to the app.

Logs were requested so we enabled NFS4 and FSAL logs on the ganesha side; however, the problem did not reproduce.    We'll retry with less logging enabled.   Any preference as to minimal logging config?  Does anyone think FSAL or NFS_V4 logs has better chance of showing something useful to help root cause this problem?

Comment 4 jeff.a.smith 2025-05-12 16:46:47 UTC
I should have mentioned that we know the EIO is generated by the client because we found the code which produced the log message, and that code path returns EIO for a write that had no error on the wire, but returned 0-bytes written: https://github.com/linux-nfs/nfsd/blob/a79be02bba5c31f967885c7f3bf3a756d77d11d9/fs/nfs/write.c#L1620

Here's the requested ceph -s output:
[ceph: root@dal1-qz2-sr5-rk025-s06 /]# ceph -s
  cluster:
    id:     12852256-21e6-11f0-8891-6cfe549bafb8
    health: HEALTH_OK

  services:
    mon: 3 daemons, quorum dal1-qz2-sr5-rk025-s06,dal2-qz2-sr2-rk089-s06,dal3-qz2-sr3-rk279-s06 (age 5d)
    mgr: dal1-qz2-sr5-rk025-s06.vdomyr(active, since 5d), standbys: dal2-qz2-sr2-rk089-s06.qougcd, dal3-qz2-sr3-rk279-s06.shdatw
    mds: 3/3 daemons up, 3 hot standby
    osd: 360 osds: 360 up (since 5d), 360 in (since 2w)

  data:
    volumes: 1/1 healthy
    pools:   6 pools, 53281 pgs
    objects: 2.10k objects, 5.3 GiB
    usage:   411 GiB used, 1.0 PiB / 1.0 PiB avail
    pgs:     53281 active+clean

  io:
    client:   3.4 MiB/s rd, 395 MiB/s wr, 36 op/s rd, 3.14k op/s wr

Comment 5 Russell Cattelan 2025-05-13 00:22:24 UTC
Created attachment 2089595 [details]
Ganesha logs with FSAL = DEBUG_FULL turned on.

Ganesha logs with FSAL = DEBUG_FULL turned on.

Comment 6 Russell Cattelan 2025-05-14 19:31:52 UTC
Created attachment 2089848 [details]
Latest debug run with EVENTS enabled

Comment 7 Russell Cattelan 2025-05-14 19:57:22 UTC
fio-3.36
Starting 4 processes
fio: io_u error on file /denali/test0: Input/output error: write offset=411146240, buflen=10240
fio: pid=20150, err=5/file:io_u.c:1896, func=io_u error, error=Input/output error

test0: (groupid=0, jobs=4): err= 5 (file:io_u.c:1896, func=io_u error, error=Input/output error): pid=20150: Wed May 14 18:06:35 2025
  mixed: IOPS=6263, BW=61.2MiB/s (64.1MB/s)(3303MiB/53997msec)
    slat (nsec): min=1646, max=7383.3k, avg=12673.05, stdev=49816.05
    clat (usec): min=3297, max=89935, avg=10194.26, stdev=5733.47
     lat (usec): min=3306, max=89938, avg=10206.93, stdev=5733.35
    clat percentiles (usec):
     |  1.00th=[ 6063],  5.00th=[ 7046], 10.00th=[ 7504], 20.00th=[ 8094],
     | 30.00th=[ 8455], 40.00th=[ 8848], 50.00th=[ 9241], 60.00th=[ 9634],
     | 70.00th=[10028], 80.00th=[10683], 90.00th=[11994], 95.00th=[14353],
     | 99.00th=[44827], 99.50th=[55837], 99.90th=[70779], 99.95th=[73925],
     | 99.99th=[82314]
    lat percentiles (usec):
     |  1.00th=[ 6063],  5.00th=[ 7046], 10.00th=[ 7504], 20.00th=[ 8094],
     | 30.00th=[ 8455], 40.00th=[ 8848], 50.00th=[ 9241], 60.00th=[ 9634],
     | 70.00th=[10028], 80.00th=[10683], 90.00th=[11994], 95.00th=[14353],
     | 99.00th=[44827], 99.50th=[55837], 99.90th=[70779], 99.95th=[73925],
     | 99.99th=[83362]
   bw (  KiB/s): min=42280, max=76720, per=99.97%, avg=62617.78, stdev=1683.07, samples=428
   iops        : min= 4228, max= 7672, avg=6261.69, stdev=168.30, samples=428
  lat (msec)   : 4=0.01%, 10=69.65%, 20=27.96%, 50=1.62%, 100=0.77%
  cpu          : usr=0.87%, sys=2.12%, ctx=296239, majf=0, minf=63
  IO depths    : 1=0.1%, 2=0.1%, 4=0.1%, 8=0.1%, 16=100.0%, 32=0.0%, >=64=0.0%
     submit    : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
     complete  : 0=0.1%, 4=100.0%, 8=0.0%, 16=0.1%, 32=0.0%, 64=0.0%, >=64=0.0%
     issued rwts: total=338212,0,0,0 short=0,0,0,0 dropped=0,0,0,0
     latency   : target=0, window=0, percentile=100.00%, depth=16

Run status group 0 (all jobs):
  MIXED: bw=61.2MiB/s (64.1MB/s), 61.2MiB/s-61.2MiB/s (64.1MB/s-64.1MB/s), io=3303MiB (3463MB), run=53997-53997msec
root@ubuntu-bx2-2x8-03:/denali#

Comment 8 jeff.a.smith 2025-05-14 23:43:31 UTC
We're currently working on getting a tcpdump wire trace of the problem.  Unfortunately, it hasn't reproduced with tcpdump running.

Comment 9 Stephen Blinick 2025-05-15 04:45:17 UTC
@msaini  -- I notice in your recreate you are running numjobs=1, and it appears the bug was opened with numbjobs=16.  Given there's a question about concurrent writes causing the issue, can we make sure the recreate efforts so far match the FIO job run in the original environment (e.g. more jobs).  Latest recreate attempt in the last comment appears to be numjobs=4.

Comment 10 jeff.a.smith 2025-05-15 14:28:03 UTC
tcpdump running inside the VSI finally captured a WRITE reply
whose status was NFS4_OK with count=0. This has only happened
once for me in about 10 reproductions.

The interesting thing is that all other WRITE call/reply pairs
that I've inspected are about 150 pkts apart in the capture
file. However, the count=0 WRITE reply (pkt 18) is the 3rd
packet after its correpsonding WRITE call (pkt 15). Pkt numbers
are from a pcap file I created by exporting the last few
hundred pkts from a 425MB pcap file.

The packet capture for other my other randwrite reproductions
show the EIO offset write reply with count=10240 which is correct.
Also interesting is that even though tcpdump reports no dropped pkts,
the pcap file does not always contain the WRITE offset that caused
fio to fail EIO. Since I'm using -s 300, the pcap file would miss
any write RPCs that started after 300 bytes in an ethernet frame, so
that's one possible explanation for "missing writes".

All this is to say that I'm not exactly trusting my tcpdump captures
collected from inside a VSI, and I'll work to get a wire trace on the
protocol node / on the metal. This is more tedious because of our
magical networking.

I'm attaching a tarball with the abbreviated capture file, fio output,
and my notes called "jeff-repro-0count_write.tar.gz"

Comment 11 jeff.a.smith 2025-05-15 14:31:24 UTC
Created attachment 2089965 [details]
capture file, fio results, and notes showing NFS4_OK write reply with count=0

tarball contains a few hundred packets exported from a 425MB tcpdump which shows a WRITE call writing 10k, and a NFS4_OK WRITE reply with count=0.   I include fio output showing file offset with EIO, and my notes.

Comment 12 Prashant Dhange 2025-05-15 19:03:18 UTC
This issue seems to occur because the ganesha server occasionally returns a zero-length write when busy, which the NFS client interprets as an error as highlighted by Jeff in earlier comment. Have we investigated ganesha server node from slow storage or high load or network congestion perspective ? While fio retries short writes, it is treating zero-length writes as errors, leading to the reported issue.

I believe the continue_on_error option in fio could help bypass these short write errors but I'm uncertain if ignoring them is the appropriate long-term solution in this context. We should see this issue only with non-direct fio runs considering the fact that nfs_writeback_result reports EIO when ganesha returns 0-bytes written. Can we try to reproduce this issue with `direct=1` fio runs ?

Comment 15 Russell Cattelan 2025-05-16 23:58:32 UTC
Created attachment 2090177 [details]
Logs from ganesha debug build 2

FIO error message with offset

io-3.36
Starting 4 processes
fio: io_u error on file /denali/test0: Input/output error: write offset=882667520, buflen=10240
fio: pid=1084, err=5/file:io_u.c:1896, func=io_u error, error=Input/output error

Comment 18 Prashant Dhange 2025-05-20 16:38:01 UTC
Hi Russell,

(In reply to Russell Cattelan from comment #15)
> Created attachment 2090177 [details]
> Logs from ganesha debug build 2
> 
> FIO error message with offset
> 
> io-3.36
> Starting 4 processes
> fio: io_u error on file /denali/test0: Input/output error: write
> offset=882667520, buflen=10240
> fio: pid=1084, err=5/file:io_u.c:1896, func=io_u error, error=Input/output
> error

Thanks for the ganesha debug logs. The ganesha logs does show nfs write op related to write call at offset 882667520 with buffer length 10240.

% grep 882667520 ganesha_mucho_logs.out
16/05/2025 21:58:55 : epoch 6827b514 : dal1-qz2-sr2-rk044-s18 : ganesha.nfsd-2[svc_28] nfs4_op_write :NFS4 :EVENT :calling write_arg 0x7f7614569800 arg_WRITE4 0x7f75ac0303a8 res_WRITE4 0x7f75ac136528 offset = 882667520  io_request = 10240  fsal_stable = 1
16/05/2025 21:58:55 : epoch 6827b514 : dal1-qz2-sr2-rk044-s18 : ganesha.nfsd-2[svc_28] ceph_fsal_write2 :FSAL :EVENT :CDBG: Calling ceph_ll_nonblocking_readv_writev for write ceph_ll_ioinfo 0x7f76141b0b88 offset 882667520 length 10240

We would like to capture the libcephfs client logs for further investigation of this issue. Kindly collect libcephfs debug logs using below steps :

[1] Enable libcephfs debug logging. On ceph cluster admin node :
# ceph config set client debug_client 20
# ceph config set client log_file /var/log/ceph/libcephfs_client.log

[2] Enable NFS4 and FSAL logs on the ganesha side

[3] Reproduce fio issue reporting "io_u error on file /denali/test0: Input/output error: write"

[4] Disable libcephfs debug logging. On ceph cluster admin node :
# ceph config rm client debug_client
# ceph config rm client log_file

[5] Collect ganesha debug logs

[6] Collect libcephfs debug logs from ganesha nodes. The /var/log/ceph/libcephfs_client.log will be generated inside the ganesha container, so collect libcephfs_client.log from every ganesha node.
- Find ganesha containers running on a ganesha node
# podman ps -a|grep ganesha

- Find libcephfs_client.log generated on every ganesha container and compress it. Upload compressed logs to BZ.
# find /var/ -iname libcephfs_client.log|xargs du -sh 
# tar -czvf /tmp/libcephfs_`hostname -s`_$(date +%d%b%y_%H-%M-%S).tgz <libcephfs_file_path_from_ganesha_container1> <libcephfs_file_path_from_ganesha_container2>
# ls -ltr /tmp
NOTE : Upload newly generated compressed file from /tmp directory prefixed by 'libcephfs_' keyword.

e.g
# podman ps -a|grep ganesha
b47b01053e12  cp.stg.icr.io/cp/ibm-ceph/ceph-8-rhel9@sha256:ca65e6bfabd1652fec495211ae72d8a4af3271bcd88ea948b623089381b982f3  -F -L STDERR -N N...  2 days ago    Up 2 days    80/tcp, 5000/tcp, 6789/tcp, 6800/tcp, 6801/tcp, 6802/tcp, 6803/tcp, 6804/tcp, 6805/tcp  ceph-f7d28fcc-310b-11f0-9028-005056bb2bda-nfs-nfsganesha-0-0-node1-xdvemh
680f0bddc833  cp.stg.icr.io/cp/ibm-ceph/ceph-8-rhel9@sha256:ca65e6bfabd1652fec495211ae72d8a4af3271bcd88ea948b623089381b982f3  -F -L STDERR -N N...  45 hours ago  Up 45 hours  80/tcp, 5000/tcp, 6789/tcp, 6800/tcp, 6801/tcp, 6802/tcp, 6803/tcp, 6804/tcp, 6805/tcp  ceph-f7d28fcc-310b-11f0-9028-005056bb2bda-nfs-ganesha1-1-0-node1-wuatin

# find /var/ -iname libcephfs_client.log|xargs du -sh
2.1G  /var/lib/containers/storage/overlay/cf94103061b9f9517ddd014d32d470f283863b3de8ef4c911b2ddb6fe6389b2b/diff/var/log/ceph/libcephfs_client.log
2.1G  /var/lib/containers/storage/overlay/cf94103061b9f9517ddd014d32d470f283863b3de8ef4c911b2ddb6fe6389b2b/merged/var/log/ceph/libcephfs_client.log
0 /var/lib/containers/storage/overlay/482dc25a8c631a6651aa2e880957f6215049624441ff2bd4351ee2e6a8c170a4/diff/var/log/ceph/libcephfs_client.log
0 /var/lib/containers/storage/overlay/482dc25a8c631a6651aa2e880957f6215049624441ff2bd4351ee2e6a8c170a4/merged/var/log/ceph/libcephfs_client.log
0 /var/log/ceph/libcephfs_client.log

# tar -czvf /tmp/libcephfs_`hostname -s`_$(date +%d%b%y_%H-%M-%S).tgz /var/lib/containers/storage/overlay/cf94103061b9f9517ddd014d32d470f283863b3de8ef4c911b2ddb6fe6389b2b/merged/var/log/ceph/libcephfs_client.log /var/lib/containers/storage/overlay/482dc25a8c631a6651aa2e880957f6215049624441ff2bd4351ee2e6a8c170a4/merged/var/log/ceph/libcephfs_client.log
tar: Removing leading `/' from member names
/var/lib/containers/storage/overlay/cf94103061b9f9517ddd014d32d470f283863b3de8ef4c911b2ddb6fe6389b2b/merged/var/log/ceph/libcephfs_client.log
tar: Removing leading `/' from hard link targets

# ls -ltr /tmp/
...
-rw-r--r--. 1 root     root     66586734 May 20 22:03 libcephfs_node1_20May25_22-03-02.tgz

Comment 22 jeff.a.smith 2025-05-22 22:15:28 UTC
Created attachment 2091162 [details]
ganesha logs from the debug image without zerocopy and FSAL async features

ganesha logs from the debug image without zerocopy and FSAL async features

VSI running fio against ganesha reproduced the EIO problem.   No wire trace was collected this time.

Comment 23 Prashant Dhange 2025-05-23 17:30:02 UTC
Hi Jeff,

(In reply to jeff.a.smith from comment #22)
> Created attachment 2091162 [details]
> ganesha logs from the debug image without zerocopy and FSAL async features
> 
> ganesha logs from the debug image without zerocopy and FSAL async features
> 
> VSI running fio against ganesha reproduced the EIO problem.   No wire trace
> was collected this time.

The ganesha logs attached in comment#22 does not contain debug logs. It only shows logs when ganesha daemon started :

May 20 14:36:01 dal3-qz2-sr3-rk247-s18 ceph-b57c0d78-e8d2-11ef-820c-7cc2554980d4-nfs-r134-50b8f3ae-99aa-46be-beb6-07ca3db878c9-0-0-dal3-qz2-sr3-rk247-s18-khutyv[3672130]: 20/05/2025 14:36:01 : epoch 682c9350 : dal3-qz2-sr3-rk247-s18 : ganesha.nfsd-2[main] nfs_Start_threads :THREAD :EVENT :General fridge was started successfully
May 20 14:36:01 dal3-qz2-sr3-rk247-s18 ceph-b57c0d78-e8d2-11ef-820c-7cc2554980d4-nfs-r134-50b8f3ae-99aa-46be-beb6-07ca3db878c9-0-0-dal3-qz2-sr3-rk247-s18-khutyv[3672130]: 20/05/2025 14:36:01 : epoch 682c9350 : dal3-qz2-sr3-rk247-s18 : ganesha.nfsd-2[main] nfs_start :NFS STARTUP :EVENT :-------------------------------------------------
May 20 14:36:01 dal3-qz2-sr3-rk247-s18 ceph-b57c0d78-e8d2-11ef-820c-7cc2554980d4-nfs-r134-50b8f3ae-99aa-46be-beb6-07ca3db878c9-0-0-dal3-qz2-sr3-rk247-s18-khutyv[3672130]: 20/05/2025 14:36:01 : epoch 682c9350 : dal3-qz2-sr3-rk247-s18 : ganesha.nfsd-2[main] nfs_start :NFS STARTUP :EVENT :             NFS SERVER INITIALIZED
May 20 14:36:01 dal3-qz2-sr3-rk247-s18 ceph-b57c0d78-e8d2-11ef-820c-7cc2554980d4-nfs-r134-50b8f3ae-99aa-46be-beb6-07ca3db878c9-0-0-dal3-qz2-sr3-rk247-s18-khutyv[3672130]: 20/05/2025 14:36:01 : epoch 682c9350 : dal3-qz2-sr3-rk247-s18 : ganesha.nfsd-2[main] nfs_start :NFS STARTUP :EVENT :-------------------------------------------------
May 20 14:36:01 dal3-qz2-sr3-rk247-s18 ceph-b57c0d78-e8d2-11ef-820c-7cc2554980d4-nfs-r134-50b8f3ae-99aa-46be-beb6-07ca3db878c9-0-0-dal3-qz2-sr3-rk247-s18-khutyv[3672130]: 20/05/2025 14:36:01 : epoch 682c9350 : dal3-qz2-sr3-rk247-s18 : ganesha.nfsd-2[reaper] nfs_try_lift_grace :STATE :EVENT :check grace:reclaim complete(0) clid count(1)
May 20 14:36:11 dal3-qz2-sr3-rk247-s18 ceph-b57c0d78-e8d2-11ef-820c-7cc2554980d4-nfs-r134-50b8f3ae-99aa-46be-beb6-07ca3db878c9-0-0-dal3-qz2-sr3-rk247-s18-khutyv[3672130]: 20/05/2025 14:36:11 : epoch 682c9350 : dal3-qz2-sr3-rk247-s18 : ganesha.nfsd-2[reaper] nfs_try_lift_grace :STATE :EVENT :check grace:reclaim complete(1) clid count(1)
May 20 14:36:11 dal3-qz2-sr3-rk247-s18 ceph-b57c0d78-e8d2-11ef-820c-7cc2554980d4-nfs-r134-50b8f3ae-99aa-46be-beb6-07ca3db878c9-0-0-dal3-qz2-sr3-rk247-s18-khutyv[3672130]: 20/05/2025 14:36:11 : epoch 682c9350 : dal3-qz2-sr3-rk247-s18 : ganesha.nfsd-2[reaper] nfs_lift_grace_locked :STATE :EVENT :NFS Server Now NOT IN GRACE

Comment 24 jeff.a.smith 2025-05-23 19:25:07 UTC
Comment on attachment 2091162 [details]
ganesha logs from the debug image without zerocopy and FSAL async features

Please ignore these logs as the container was not running the latest debug code from Frank.

Comment 25 Russell Cattelan 2025-05-23 20:37:51 UTC
Created attachment 2091331 [details]
Latest run with async turned off.

fio include.fio --filename="$(pwd)/test0"
test0: (g=0): rw=randwrite, bs=(R) 10.0KiB-10.0KiB, (W) 10.0KiB-10.0KiB, (T) 10.0KiB-10.0KiB, ioengine=libaio, iodepth=16
...
fio-3.36
Starting 4 processes
fio: io_u error on file /denali/test0: Input/output error: write offset=942694400, buflen=10240
fio: pid=19578, err=5/file:io_u.c:1896, func=io_u error, error=Input/output error

test0: (groupid=0, jobs=4): err= 5 (file:io_u.c:1896, func=io_u error, error=Input/output error): pid=19578: Fri May 23 16:36:41 2025

Comment 33 jeff.a.smith 2025-05-27 15:24:10 UTC
This morning I verified that QOS feature is configured and enabled in a11 (the CEPH cluster in our ibmcloud env reproducing EIO): 

[ceph: root@dal1-qz2-sr2-rk044-s18 /]# ceph nfs cluster qos get r134-0803038e-8f26-4335-a1c1-a75707fce95d
{
  "combined_rw_bw_control": true,
  "enable_bw_control": true,
  "enable_iops_control": true,
  "enable_qos": true,
  "max_export_combined_bw": "1.1GB",
  "max_export_iops": 35000,
  "qos_type": "PerShare"
}

Comment 46 errata-xmlrpc 2025-06-26 12:31:40 UTC
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory (Important: Red Hat Ceph Storage 8.1 security, bug fix, and enhancement updates), and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHSA-2025:9775

Comment 47 konstas 2025-06-28 03:20:01 UTC
>Since the problem described in this bug report should be
resolved in a recent advisory,
>it has been closed with a
resolution of ERRATA.
>
>For information on the advisory (Important: Red Hat Ceph Storage 8.1 security, bug fix, and enhancement updates), and where to find the updated
files, follow the link below.
>
>If the solution does not work for you, open a new bug report.
>https://access.redhat.com/errata/RHSA-2025:9775 https://rbtv77apk.app
>
>Thanks Fellow!


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