[Fwd: Cron <root at app4> /usr/local/bin/restart-memhogs.sh]
mmcgrath at redhat.com
Sun Dec 2 18:28:50 UTC 2007
Toshio Kuratomi wrote:
> Looks like smolt is hitting the limit of 700MB pretty easily. I was
> going to up the limit but I found that smolt was timing out even when
> running with less than 700MB of memory. Restarting the server stopped
> that from happening.
> So one possibility is that once smolt allocates so much memory, some
> portion of what it does takes too much time to scan through that
> memory. If that's so we actually want to decrease the amount of
> memory smolt can allocate before being restarted.
> Another possibility is that the smolt server is getting stuck
> somewhere in processing and that's preventing either the smolt server
> or the database from processing future sendProfile's correctly. In
> this scenario, restarting the server due to memory usage is fixing
> this as a side effect. We really want to find and fix the bug if this
> is the case.
Smolt is running heaver than normal right now as all of the Fedora
clients are doing their monthly checkin over the next few days.
Fortunately smolt hasn't really been restarted until this heavier
period. Unfortunately now that we're seeing a bunch of check-ins we'll
have to take a closer look into what is going on. It should be pretty
easy for us to verify on another machine.
More information about the Fedora-infrastructure-list