[Bugs] [Bug 1313567] flooding of "dict is NULL" logging

bugzilla at redhat.com bugzilla at redhat.com
Thu Feb 7 03:01:01 UTC 2019


https://bugzilla.redhat.com/show_bug.cgi?id=1313567



--- Comment #20 from Nithya Balachandran <nbalacha at redhat.com> ---
(In reply to Artem Russakovskii from comment #19)
> (In reply to Nithya Balachandran from comment #18)
> 
> Before I make this change, what is the expected effect on performance on a
> busy system with lots of small and large files (millions)?

This turns off a feature that was introduced in 5.3 to clear unused inodes from
memory for the fuse client gluster process and prevent the memory usage from
rising. Turning it off would cause the client to behave the same as in the
earlier releases. 

> 
> Also, is it possible to set the lru-limit to 0 via gluster volume options
> (network.inode-lru-limit?) or will it be sufficiently different from using
> the fuse mount option?

Not at the moment. This is a mount option for the gluster fuse client. This is
not the same as the network.inode-lru-limit option which controls the inodes in
the brick processes.
> 
> Do gluster docs have clear instructions on how to capture a core dump
> because I haven't found any other than the instructions to provide one (but
> not how) in case of a crash?


To configure coredumps, you need to specify the size and location. The
following link describes it well:
https://stackoverflow.com/questions/17965/how-to-generate-a-core-dump-in-linux-on-a-segmentation-fault

Make sure you have sufficient space in the location for a large coredump (it
can run into several GBs)

-- 
You are receiving this mail because:
You are on the CC list for the bug.


More information about the Bugs mailing list