Thanks David, At least I'm not crazy. At least. What feels wrong for me is that the data behind the systems didn't change, so it's not like suddenly Plack had much more stuff to hold in cache. I didn't consider that a new version of plack/starman could be in cause. plack.psgi hasn't changed much in the past year. x2 or x3 is a lot, especially when you do that times the number of worker, times the number of clients... Philippe Blouin, Directeur de la technologie Tél. : (833) 465-4276, poste 230 philippe.blouin@inLibro.com inLibro | pour esprit libre | www.inLibro.com <http://www.inLibro.com> On 2026-02-18 23:03, David Cook wrote:
Hi Philippe,
That is an interesting observation. I’ve just compared a number of 23.11 and 24.11 instances, and yeah I’d say most of the 23.11 are hovering around 300MB. There are some 400-600 but mostly 300ish.
If I look at 24.11… actually most of them are around 300MB, but there are some that are more like 800-1100MB.
But that being said… that’s VIRT memory rather than RES memory. The RES memory is actually in the 300-400 mark as we’d expect.
If I understand correctly, that means that these processes have allocated large amounts of memory, but are currently only using the normal amount of memory. That suggests that they had a previous request where they had to allocate more memory for that request.
Now Perl can re-use memory it has allocated, but it cannot return memory to the system until the process has ended. So it has likely internally freed up that memory, but it still has an allocation on all that memory space.
So… I’d say there isn’t a memory leak since the RES is low, but it sure makes you wonder what causes the high VIRT.
In my case, my processes with 1100MB have been hit by bots going to opac-search.pl, opac-detail.pl, and opac-image.pl. That said, cover_images can only house files up to 16MB, so it’s not like it’s serving huge files.
It’s certainly odd. Nothing in the logs really stands out to me. One thought I did have was that maybe a bot sent a very large POST request or something, but nothing really stands out to me.
Curious…
David Cook
Senior Software Engineer
Prosentient Systems
Suite 7.03
6a Glen St
Milsons Point NSW 2061
Australia
Office: 02 9212 0899
*From:*Koha-devel <koha-devel-bounces@lists.koha-community.org> *On Behalf Of *Philippe Blouin via Koha-devel *Sent:* Thursday, 19 February 2026 2:18 AM *To:* koha-devel@lists.koha-community.org *Subject:* [Koha-devel] Plack memory usage since 25
Hi all!,
Has anything changed in the code between 24.05 and 25.05 that could cause Plack to use a lot more memory ?
My starman processes use to top at around 300M when using 24.05. Now with exactly the same settings, these processes have all inflated and some are now topping 900 (!!!!) M.
Any suggestion, idea ?
Much appreciated,
--
Philippe Blouin, Directeur de la technologie
Tél. : (833) 465-4276, poste 230 philippe.blouin@inLibro.com <mailto:philippe.blouin@inLibro.com>
inLibro| pour esprit libre |www.inLibro.com <http://www.inLibro.com>