Re: [HACKERS] Re: [COMMITTERS] pgsql: Make new crash restart test a bit more robust.

2017-09-21 Thread Thomas Munro
On Wed, Sep 20, 2017 at 4:42 PM, Andres Freund wrote: > On 2017-09-19 19:00:38 -0700, Andres Freund wrote: >> Given this fact pattern, I'll allow the case without a received error >> message in the recovery test. Objections? > > Hearing none. Pushed. > > While debugging this, I've also introduced

Re: [HACKERS] Re: [COMMITTERS] pgsql: Make new crash restart test a bit more robust.

2017-09-19 Thread Andres Freund
On 2017-09-19 19:00:38 -0700, Andres Freund wrote: > Given this fact pattern, I'll allow the case without a received error > message in the recovery test. Objections? Hearing none. Pushed. While debugging this, I've also introduced a pump wrapper so that we now get: ok 4 - exactly one process kil

Re: [COMMITTERS] pgsql: Make new crash restart test a bit more robust.

2017-09-19 Thread Andres Freund
On 2017-09-19 18:06:29 -0700, Andres Freund wrote: > On 2017-09-19 16:46:58 -0400, Tom Lane wrote: > > Have we forgotten an fflush() or something? > > After hacking a fix for my previous theory, I started adding strace into > the mix, to verify this. Takes longer to reproduce, but after filtering

Re: [COMMITTERS] pgsql: Make new crash restart test a bit more robust.

2017-09-19 Thread Andres Freund
On 2017-09-19 16:46:58 -0400, Tom Lane wrote: > Andres Freund writes: > > So this is geniuinely interesting. When the machine is really loaded (as > > in 6 animals running on a vm at the same time, incuding valgrind), psql > > sometimes doesn't get the WARNING message from a shutdown. Instead it >

Re: [COMMITTERS] pgsql: Make new crash restart test a bit more robust.

2017-09-19 Thread Andres Freund
On 2017-09-19 13:53:18 -0700, Andres Freund wrote: > On 2017-09-19 16:46:58 -0400, Tom Lane wrote: > > Have we forgotten an fflush() or something? > > > > Also, maybe problem is on client side. I vaguely recall a libpq bug > > wherein it would complain about socket EOF even though data remained >

Re: [COMMITTERS] pgsql: Make new crash restart test a bit more robust.

2017-09-19 Thread Tom Lane
Andres Freund writes: > I'll see if I can spot the bug in an hour or two. If not I'll make the > test temporarily accept both outputs while investigating? Seems reasonable. regards, tom lane -- Sent via pgsql-committers mailing list (pgsql-committers@postgresql.org) To

Re: [COMMITTERS] pgsql: Make new crash restart test a bit more robust.

2017-09-19 Thread Andres Freund
On 2017-09-19 16:46:58 -0400, Tom Lane wrote: > Andres Freund writes: > > So this is geniuinely interesting. When the machine is really loaded (as > > in 6 animals running on a vm at the same time, incuding valgrind), psql > > sometimes doesn't get the WARNING message from a shutdown. Instead it >

Re: [COMMITTERS] pgsql: Make new crash restart test a bit more robust.

2017-09-19 Thread Tom Lane
Andres Freund writes: > So this is geniuinely interesting. When the machine is really loaded (as > in 6 animals running on a vm at the same time, incuding valgrind), psql > sometimes doesn't get the WARNING message from a shutdown. Instead it > gets > # psql::3: server closed the connection unexpe

Re: [COMMITTERS] pgsql: Make new crash restart test a bit more robust.

2017-09-19 Thread Andres Freund
On 2017-09-19 15:24:49 -0400, Tom Lane wrote: > Andres Freund writes: > > Checkining on calliphoridae why that's not sufficient - the machine's > > busy, so the build & test will take a bit. > > FWIW, prairiedog got through the recovery tests this time --- run's > still going though. So this is

Re: [COMMITTERS] pgsql: Make new crash restart test a bit more robust.

2017-09-19 Thread Tom Lane
Andres Freund writes: > Checkining on calliphoridae why that's not sufficient - the machine's > busy, so the build & test will take a bit. FWIW, prairiedog got through the recovery tests this time --- run's still going though. regards, tom lane -- Sent via pgsql-commit

Re: [COMMITTERS] pgsql: Make new crash restart test a bit more robust.

2017-09-19 Thread Andres Freund
On 2017-09-19 17:40:20 +, Andres Freund wrote: > Make new crash restart test a bit more robust. > > Add timeouts in case psql doesn't deliver the expected output, and try > to cause the monitoring psql to be fully connected to a backend. This > isn't necessarily everything needed, but at leas

[COMMITTERS] pgsql: Make new crash restart test a bit more robust.

2017-09-19 Thread Andres Freund
Make new crash restart test a bit more robust. Add timeouts in case psql doesn't deliver the expected output, and try to cause the monitoring psql to be fully connected to a backend. This isn't necessarily everything needed, but at least the timeouts should reduce the pain for buildfarm owners.