[squid-users] Significant memory leak with version 5.x (not with 4.17)

[Praveen Ponakanti]

Hi Alex,

I have attached a log with about 24 hours of memory stats from our squid
proxy running version 5.3. The memory usage increased during the first
11-12 hours of the log and then flatted out. The same pattern repeats
itself each day with about an additional 2-3G being used up. Let me know if
you need any additional logs. I will take another shot at the test setup
with Valgrind later on. Happy holidays!

https://bugs.squid-cache.org/show_bug.cgi?id=5132

Thanks
Praveen

On Thu, Dec 23, 2021 at 6:23 AM Alex Rousskov <
rousskov at measurement-factory.com> wrote:

> On 12/23/21 1:58 AM, Praveen Ponakanti wrote:
> > On Wed, Dec 22, 2021 at 6:45 AM Alex Rousskov wrote:
> >
> >     On 12/21/21 7:48 PM, Praveen Ponakanti wrote:
> >
> >     > We are running the squid proxy for servicing outbound HTTP quests
> from
> >     > our network and have observed a significant memory leak with 5.x
> >     > versions. While there are several discussions about memory leaks
> with
> >     > recent versions, just wanted to list out what we have observed in
> case
> >     > this is an unknown leak.
> >
> >     I recommend sharing a log with 48+ hourly mgr:mem snapshots. These
> >     snapshots help compare your leak with others we know about and may
> help
> >     isolate some of the memory leaks:
> >     https://bugs.squid-cache.org/show_bug.cgi?id=5132#c8
>
> > I will continue collecting these stats hourly.
>
> I will analyze those logs once you share them. Comparing two snapshots
> from different Squid versions/instances may be useful, but it is much
> more time consuming, and I cannot offer to do that right now.
>
>
> > The output appears to be quite large
>
> > ... where do I send the files with the memory logs?
>
> The scripts that track memory usage changes across hourly snapshots do
> not care about the log size. However, you may want to share a link to
> the (compressed) log rather than sharing the log itself. The best place
> to share that info, IMO, is the bug report I referenced above, but you
> can post here, of course.
>
>
> > I might have to fix how it is launched from
> > the docker entrypoint script in my test env.
>
> FWIW, here is one of the valgrind commands I use (with paths stripped):
>
> > valgrind -v --trace-children=no --child-silent-after-fork=yes
> --num-callers=50 --log-file=valgrind-%p.log --time-stamp=yes
> --leak-check=full --leak-resolution=high --show-reachable=no
> --track-origins=no --gen-suppressions=all --suppressions=valgrind.supp
> squid -f squid.conf ...
>
> Valgrind reports useful leak info after Squid exits.
>
> I do not have updated/polished Valgrind suppressions, but I can share
> what I use:
> https://gist.github.com/rousskov/a1f503981b4a3832ab44221b0b615b5a
>
> I do not use gdb/valgrind monitor so I cannot help you there.
>
>
> HTH,
>
> Alex.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.squid-cache.org/pipermail/squid-users/attachments/20211223/a5f48b3a/attachment.htm>