On Fri, Mar 24, 2017 at 06:11:01PM +0100, Corinna Vinschen wrote: > - cygserver is using a defined number of threads in a thread pool for > application requests. Every request is added to a request submission > queue and handled by the next free thread in the pool. > > The default number of threads in the pool is 10. Each wait for a > semaphore is blocking one thread. If more than the number of threads > in the pool are supposed to wait on a semaphore the pool starves.
Interesting. I can confirm that, without updating software, "cygserver -r 40" fixes both my self-contained test and my PostgreSQL test case. Folks can use that workaround in released-version installations. > So what I did now is to allow cygserver to raise the number of worker > threads on demand. That is, if a request is in the queue and all > worker threads are busy, just create a new one. > > There's no way yet to drop threads again, but this should be a minor > problem in scenarions which really have a lot of contention. Agreed. This is nicer. > I pushed a patchset now, and uploaded new developer snapshots for > testing to https://cygwin.com/snapshots/ > Please give it a try Self-contained test case results look good: cygwin-20170321.tar.xz "cygserver -r40": ok cygwin-20170324.tar.xz "cygserver -r40": ok cygwin-20170321.tar.xz "cygserver -r10": freezes (expected) cygwin-20170324.tar.xz "cygserver -r10": ok; cygserver output concludes with "cygserver: All threads busy, added one (now 21)". I then tried my PostgreSQL test case ("pgbench -i -s 50" once to setup, then "pgbench -S -j2 -c16 -T900 -P5" to test): cygwin-20170321.tar.xz "cygserver -r40": ok for >3600s cygwin-20170324.tar.xz "cygserver -r40": limited freeze in <1000s; no cygserver output cygwin-20170321.tar.xz "cygserver -r10": classic freeze in <1000s (expected) cygwin-20170324.tar.xz "cygserver -r10": limited freeze in <1000s; no cygserver output for most of the run, then output concluding with "cygserver: All threads busy, added one (now 15)" just before the freeze I call the cygwin-20170324 freezes "limited" because the symptoms differ from the classic freeze I described upthread. "strace /bin/true" and "cat /proc/sysvipc/sem" do not hang, but every PostgreSQL backend process is stuck waiting on a synchronization primitive. I can distill another self-contained test case for the limited freeze seen in cygwin-20170324, but that make take awhile. I'm sending this early report so you're aware of the possible regression in cygwin-20170324. Thanks, nm -- Problem reports: http://cygwin.com/problems.html FAQ: http://cygwin.com/faq/ Documentation: http://cygwin.com/docs.html Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple