bug#58631: [Shepherd] Indefinite heap growth (memory leak)

2022-11-11 Thread Ludovic Courtès
Hi! Ludovic Courtès skribis: > Bayfront.guix has been running it for more than 12h and heap size hasn’t > changed in that time span. Heap size has remained unchanged for the past 48h. Berlin.guix was reconfigured & rebooted 12h and shepherd’s heap size has remained unchanged. Officially closi

bug#58631: [Shepherd] Indefinite heap growth (memory leak)

2022-11-10 Thread Ludovic Courtès
With the Fibers fix for , backported in Guix commit d9ca9cdd01bf1097343a047b51a1392131c7cf58, this shepherd memory leak seems to be finally fixed (details in the issue above). Bayfront.guix has been running it for more than 12h and heap size hasn’t change

bug#58631: [Shepherd] Indefinite heap growth (memory leak)

2022-11-07 Thread Ludovic Courtès
Ludovic Courtès skribis: > Ludovic Courtès skribis: > >> Ludovic Courtès skribis: >> >>> The attached Fibers program illustrates the problem: heap grows even >>> though it’s not supposed to. >> >> The saga continues at . > > … and then at

bug#58631: [Shepherd] Indefinite heap growth (memory leak)

2022-11-06 Thread Ludovic Courtès
Ludovic Courtès skribis: > Ludovic Courtès skribis: > >> The attached Fibers program illustrates the problem: heap grows even >> though it’s not supposed to. > > The saga continues at . … and then at , which appears t

bug#58631: [Shepherd] Indefinite heap growth (memory leak)

2022-10-30 Thread Ludovic Courtès
Ludovic Courtès skribis: > The attached Fibers program illustrates the problem: heap grows even > though it’s not supposed to. The saga continues at .

bug#58631: [Shepherd] Indefinite heap growth (memory leak)

2022-10-29 Thread Maxime Devos
On 29-10-2022 12:01, Ludovic Courtès wrote: The attached Fibers program illustrates the problem: heap grows even though it’s not supposed to. If I add (gc) before the (pk 'heap-size ...), I can't reproduce: ;;; (heap-size 12742656) ;;; (heap-size 12808192) ;;; (heap-size 12808192) ;;; (he

bug#58631: [Shepherd] Indefinite heap growth (memory leak)

2022-10-29 Thread Ludovic Courtès
The attached Fibers program illustrates the problem: heap grows even though it’s not supposed to. Ludo’. ;; https://issues.guix.gnu.org/58631 (use-modules (fibers) (fibers channels) (ice-9 rdelim) (statprof)) (run-fibers (lambda () (define channel

bug#58631: [Shepherd] Indefinite heap growth (memory leak)

2022-10-23 Thread Ludovic Courtès
Liliana Marie Prikler skribis: > Am Donnerstag, dem 20.10.2022 um 11:29 +0200 schrieb Ludovic Courtès: >> Logging as performed by ‘%service-file-logger’ is quite GC-intensive >> (but shouldn’t be leaky!); this bit: >> >> (let ((prefix (strftime default-logfile-date-format >>    

bug#58631: [Shepherd] Indefinite heap growth (memory leak)

2022-10-23 Thread Ludovic Courtès
Liliana Marie Prikler skribis: > Am Donnerstag, dem 20.10.2022 um 11:29 +0200 schrieb Ludovic Courtès: >> Logging as performed by ‘%service-file-logger’ is quite GC-intensive >> (but shouldn’t be leaky!); this bit: >> >> (let ((prefix (strftime default-logfile-date-format >>    

bug#58631: [Shepherd] Indefinite heap growth (memory leak)

2022-10-23 Thread Ludovic Courtès
An update: I can reproduce in a VM running a slightly simplified version of ‘hydra/berlin.scm’, with loop running: while true; do wget -qO/dev/null http://localhost:3000/nix-cache-info; done to trigger ‘guix publish’ logging. Better, I can mostly reproduce the issue with the attached config fi

bug#58631: [Shepherd] Indefinite heap growth (memory leak)

2022-10-20 Thread Ludovic Courtès
Joshua Branson skribis: > So, this is a hack, but perhaps berlin could start to use endlessh on > port 22, and then the real ssh port could be on another standard port. We should do that, I agree; we should still fix the actual bug though! Ludo’.

bug#58631: [Shepherd] Indefinite heap growth (memory leak)

2022-10-20 Thread Joshua Branson via Bug reports for GNU Guix
Ludovic Courtès writes: > Indeed, the core dump weighs 13 GiB, so it looks like a memory leak (too > bad ‘info proc stat’ in GDB doesn’t work). > > I have not observed it on any other machine. The difference between > berlin and machines I have access to is that berlin is being hammered on > por