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

[Praveen Ponakanti]

I cant make any changes to our prod squids this week. I have a squid
instance (5.3v) in a test env but could not reproduce the leak by starting
& stopping traffic with a bulk http req generator (wrk). Was able to send
175k rps @ 20k concurrent sessions (each doing a get on a 1KB object)
through the 30-worker squid. This initially caused a 3G increase in memory
usage and then flattened out after stopping the requests. If I restart the
bulk reqs, the memory usage only goes up ~0.5GB and then drops back down.
Live traffic is probably exercising a different code path within squid's
memory pools.

On Mon, Dec 27, 2021 at 2:26 AM Lukáš Loučanský <loucansky.lukas at kjj.cz>
wrote:

> After one day of running without clients my squid memory is stable
>
> 29345 proxy     20   0  171348 122360  14732 S   0.0   0.7   0:25.96
> (squid-1) --kid squid-1 -YC -f /etc/squid5/squid.conf
> 29343 root      20   0  133712  79264   9284 S   0.0   0.5   0:00.00
> /usr/sbin/squid -YC -f /etc/squid5/squid.conf
>
> Storage Mem size:	3944 KB
> Storage Mem capacity:	 0.2% used, 99.8% free
> 	
> Maximum Resident Size: 489440 KB
> Page faults with physical i/o: 0
> Memory accounted for:
> Total accounted:        15741 KB
> memPoolAlloc calls:   1061495
> memPoolFree calls:    1071691
>
> Total allocated 15741	kB
>
> So this does not seem to be the problem...
> L
>
> Dne 26.12.2021 v 10:02 Lukáš Loučanský napsal(a):
>
> ok - as it seems my squid quacked on low memory again today -
>
> Dec 26 00:04:25 gw (squid-1): FATAL: Too many queued store_id requests;
> see on-persistent-overload.#012    current master transaction:
> master4629331
> Dec 26 00:04:28 gw squid[15485]: Squid Parent: squid-1 process 15487
> exited with status 1
> Dec 26 00:04:28 gw squid[15485]: Squid Parent: (squid-1) process 28375
> started
>
> 2021/12/26 00:01:20 kid1| helperOpenServers: Starting 5/64
> 'storeid_file_rewrite' processes
> 2021/12/26 00:01:20 kid1| ipcCreate: fork: (12) Cannot allocate memory
> 2021/12/26 00:01:20 kid1| WARNING: Cannot run
> '/lib/squid5/storeid_file_rewrite' process.
> 2021/12/26 00:01:20 kid1| ipcCreate: fork: (12) Cannot allocate memory
>
> I'm going to reroute my clients (which are on their days off anyway) to
> direct connections and run it "dry" - on it's own. But I'm not able to to
> test it before "lack of memory issues occur" - because my clients are
> offline. So I'll watch squid for it's own memory consuption. It's all I can
> do right now - my squid already restarted and it's memory has been freed -
> so I think just now I have no power to fill it up again :-]
>
> L
>
> Dne 26.12.2021 v 7:41 Amos Jeffries napsal(a):
>
>
> If possible can one of you run a Squid to get this behaviour, then stop
> new clients connecting to it before lack of memory issues occur and see if
> the memory usage disappears or reduces after a 24-48hr wait.
>
> A series of regular mempools report dumps from across the test may help
> Alex or whoever works on the bug eliminate further which cache and client
> related things are releasing properly.
>
>
> Amos
>
> _______________________________________________
> squid-users mailing list
> squid-users at lists.squid-cache.org
> http://lists.squid-cache.org/listinfo/squid-users
>
>
>
>
> <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=emailclient> Bez
> virů. www.avast.com
> <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=emailclient>
> <#m_9217020348889694418_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
> _______________________________________________
> squid-users mailing list
> squid-users at lists.squid-cache.org
> http://lists.squid-cache.org/listinfo/squid-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.squid-cache.org/pipermail/squid-users/attachments/20211227/124e5f23/attachment.htm>