- 05 May, 2013 13 commits
-
-
Geoff Simmons authored
-
Geoff Simmons authored
-
Geoff Simmons authored
-
Geoff Simmons authored
-
Geoff Simmons authored
-
Geoff Simmons authored
-
Geoff Simmons authored
- more carefully counting the running and exited workers
-
Geoff Simmons authored
-
Geoff Simmons authored
- "< max threads running" is a warning, not alert
-
Geoff Simmons authored
- restored hash entry check against fd, but not if stale (-1) - removed membar after client bit not set (didn't help) - intial data send fail (before reconnect) is a warning (not alert)
-
Geoff Simmons authored
set -- false fd's tend to come after this, maybe data are stale - dump the data table before abort
-
Geoff Simmons authored
-
Geoff Simmons authored
-
- 25 Feb, 2013 3 commits
-
-
Geoff Simmons authored
-
Geoff Simmons authored
-
Geoff Simmons authored
so that the old child process shuts down without delay
-
- 24 Feb, 2013 2 commits
-
-
Geoff Simmons authored
- adopting varnishd.c's hack of calling nm -an to get symbolic names for stack traces
-
Geoff Simmons authored
- no connection pooling, always one connection per worker
-
- 21 Feb, 2013 2 commits
-
-
Geoff Simmons authored
-
Geoff Simmons authored
-
- 20 Feb, 2013 7 commits
-
-
Geoff Simmons authored
-
Geoff Simmons authored
- monitor stats emitted one last time before abort
-
Geoff Simmons authored
because RPM is uncommonly stupid
-
Geoff Simmons authored
-
Geoff Simmons authored
-
Geoff Simmons authored
-
Geoffrey.Simmons@uplex.de authored
-
- 19 Feb, 2013 13 commits
-
-
Geoff Simmons authored
-
Geoff Simmons authored
-
Geoff Simmons authored
-
Geoff Simmons authored
-
Geoff Simmons authored
-
Geoff Simmons authored
-
Geoff Simmons authored
-
Geoff Simmons authored
-
Geoff Simmons authored
(trying to hint the compiler to use constant bit mask in hash)
-
Geoff Simmons authored
-
Geoff Simmons authored
-
Geoff Simmons authored
- reworked config params: - maxopen.scale can be an power of 2 - maxdone, maxdata and qlen.goal not powers of 2
-
Geoff Simmons authored
-