Upstream patch : https://review.gluster.org/23285
This bug is fixed in : glusterfs-3.12.2-47.5
Build : glusterfs-3.12.2-47.5 As the fix is to free memory when Interrupt happens. Utilized the script mentioned in https://bugzilla.redhat.com/show_bug.cgi?id=1728047#c0 1. Created a 1X3 replica volume, started it and mounted it 2. Compiled open_and_sleep.c from "rhs-glusterfs/tests/features" downstream source 3. Ran i=1; while :; do echo -en "\r$i "; (./open_and_sleep a1 | { sleep 0.1; xargs -n1 kill -INT; }); i=$(($i+1)); done 4. Took statedumps for every 1 minute and top output of glusterfs process. ############### 2019-09-12 20:58:01.491998 ####################### top - 20:58:01 up 1 day, 3:59, 3 users, load average: 0.00, 0.01, 0.05 Tasks: 1 total, 0 running, 1 sleeping, 0 stopped, 0 zombie %Cpu(s): 0.0 us, 0.0 sy, 0.0 ni,100.0 id, 0.0 wa, 0.0 hi, 0.0 si, 0.0 st KiB Mem : 16266308 total, 14989736 free, 288680 used, 987892 buff/cache KiB Swap: 5242876 total, 5242876 free, 0 used. 15672048 avail Mem PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 20506 root 20 0 690308 14784 3912 S 0.0 0.1 0:30.05 glusterfs ############### 2019-09-13 11:15:07.661709 ####################### top - 11:15:07 up 1 day, 18:17, 1 user, load average: 0.33, 0.14, 0.08 Tasks: 1 total, 0 running, 1 sleeping, 0 stopped, 0 zombie %Cpu(s): 4.9 us, 9.8 sy, 0.0 ni, 85.2 id, 0.0 wa, 0.0 hi, 0.0 si, 0.0 st KiB Mem : 16266308 total, 14946580 free, 285168 used, 1034560 buff/cache KiB Swap: 5242876 total, 5242876 free, 0 used. 15630936 avail Mem PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 20506 root 20 0 755844 14816 3912 S 0.0 0.1 5:28.54 glusterfs #################### For 4,69,110 iterations, seen a slight increase of 32kB which is ran for 15 hours. No increase in size and num_allocs seen for "gf_fuse_mt_iov_base" and fix is working as expected. Scenario 2: 1. Created 1X3 volume, started and mounted it. 2. Used the compiled code to run from 4 screens 3. Ran the script from 4 screens for 70,000 iterations 4. Took statdumps for every 10 mins and top output of glusterfs process No memory increase of glusterfs process No increase in size and num_allocs seen for "gf_fuse_mt_iov_base" and fix is working as expected. Moving this bug to verified.
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, 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/RHBA-2019:2812
*** Bug 1786327 has been marked as a duplicate of this bug. ***