Daniel Farina
Daniel Farina
Could you do me a favor and clean up the commit history however you see fit?
That's weird. Although, it looks like there may be a timeout at the root of this?
No, but I do run quite a few computers, and this particular problem has a telltale symptom, so I can survey for it now and then. If I see it...
Got a live one here. strace looks a bit like this: ``` setsockopt(13, SOL_SOCKET, SO_OOBINLINE, [1], 4) = 0 epoll_ctl(3, EPOLL_CTL_ADD, 13, {EPOLLIN|EPOLLOUT, {u32=318296336, u64=94309210181904}}) = 0 epoll_wait(3, [{EPOLLIN|EPOLLOUT|EPOLLERR|EPOLLHUP, {u32=318296336,...
I've recently upgraded syslog-ng to a recent version, but it takes a time to produce (or fail to produce) any more instances of this, so, current state for me is...
The SDK works pretty well, so please do not let it slide into disrepair, which is probably the most important thing. Thanks for taking care of it.
@petere very long time no see, let me know if you need me to drop in with gdb and check upon anything. I figured you might know of an errata...
Well, what would you like though? Is there some data structure that might be useful to figure why pgbouncer is leaving stuff in sv_active even though it doesn't seem any...
Are we sure binary size is the dominant factor in startup time? I am at least equally suspicious about library initialization. mmap and faulting a few pages for instructions for...
Nice analysis. But it raises one last question: why did wal-g start so much faster ages ago? (tracing the main function does not catch library initialization, right?)