[Bugs] [Bug 1623107] FUSE client's memory leak

[bugzilla at redhat.com]

https://bugzilla.redhat.com/show_bug.cgi?id=1623107



--- Comment #38 from Znamensky Pavel <kompastver at gmail.com> ---
(In reply to Amar Tumballi from comment #37)
> While a release with this patch merged/tested is another 50days away, we
> surely would like to reduce the CPU load you see too. Whenever you get time,
> if you can capture CPU info with below tool "perf record -ag
> --call-graph=dwarf -o perf.data -p <pid of glusterfs process>", and then see
> "perf report" to see what actually caused the CPU usage, it will help us to
> resolve that too.

I'm sorry for the delay. I did `find` on a directory with ~ 200_000 files on v6
with lru-limit, without lru-limit and on v4.1.
`ps aux` shows this:

v6 with lru-limit = 10_000
USER       PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
root     30968  2.2  0.3 637936 54056 ?        Ssl  18:15   0:11
/usr/sbin/glusterfs --read-only --lru-limit=10000 --process-name fuse
--volfile-server=srv --volfile-id=/st1 /mnt/st1

v6 without lru-limit
USER       PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
root     31193  2.0  2.0 834544 330680 ?       Ssl  18:32   0:09
/usr/sbin/glusterfs --read-only --process-name fuse --volfile-server=srv
--volfile-id=/st1 /mnt/st1

v4.1
USER       PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
root     31416  1.3  2.0 817704 344996 ?       Ssl  18:41   0:06
/usr/sbin/glusterfs --read-only --process-name fuse --volfile-server=srv
--volfile-id=/st1 /mnt/st1

Also, I've got `perf` reports, but I'm afraid I can't correctly understand it.
Unfortunately, these reports have sensitive information, so I can't attach them
to this issue. Nevertheless, if it would be helpful, I could send them directly
to you.

(In reply to Amar Tumballi from comment #37)
> Also note, lru-limit=10000 while many files are accessed may not be a good
> value. I recommend something like 64k at least. But well, it depends on your
> memory needs too. So, if you can give 512MB - 1GB RAM for glusterfs, its
> better at least for performance.

Thanks for the advice!

Also, I suppose I should create a new issue for CPU problem, should I?

-- 
You are receiving this mail because:
You are on the CC list for the bug.