You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In issue #1422, user @nu11gravity ran out of memory because his system had a 10 MB default stack size and 256MB of memory, resulting in hitting a resource allocation limit in a pthread_create call since the default I/O pool size is 32.
A workaround is to use --io-threads to specify a smaller I/O pool size.
Our blocker pool threads do not need much stack space. They don't recurse, the deepest we go is when printing a backtrace when a guarantee fails.
The default stack size on the system I'm using right now is 2 MB. There's no need for use to spend 64 MB of memory on these I/O pool threads. We could set a pthread attribute to make their stacks much smaller. 128 KB would be fine, that's what our coroutine stack size is.
edit: working on this in branch sam_1425 off v1.9.x.
The text was updated successfully, but these errors were encountered:
In issue #1422, user @nu11gravity ran out of memory because his system had a 10 MB default stack size and 256MB of memory, resulting in hitting a resource allocation limit in a pthread_create call since the default I/O pool size is 32.
A workaround is to use
--io-threads
to specify a smaller I/O pool size.Our blocker pool threads do not need much stack space. They don't recurse, the deepest we go is when printing a backtrace when a guarantee fails.
The default stack size on the system I'm using right now is 2 MB. There's no need for use to spend 64 MB of memory on these I/O pool threads. We could set a pthread attribute to make their stacks much smaller. 128 KB would be fine, that's what our coroutine stack size is.
edit: working on this in branch sam_1425 off v1.9.x.
The text was updated successfully, but these errors were encountered: