onlyjob
On slow HDD I've identified heavy directories using the following command for D in $(find -maxdepth 1 -type d -printf '%P\n' | sort); do printf "\t$D\t"; tree $D | wc...
I find performance of `mfsdirinfo -p` to be adequate, no complaints here. RAM is OK. No hardware errors reported, no problems observed. Not even on secondary chunkserver serving different label...
Thank you, Agata.
It looks like commit abdd56e helped to some extent but not entirely. We just had another manifestation of this problem, only for about 10 minutes but quite prominent with I/O...
Still having this issue on 3.0.113. It manifests prominently on chunkserver restart...
I think the bottleneck causing this issue is CPU on the chunkserver. Master CPU usage during scanning did not freeze clients in my case. Upgrading CPU on the chunkserver with...
I just experienced this problem again on 3.0.116. Context: partial loss of redundancy (endangered chunks) in unrelated storage class; chunkserver with the only copy of data is reloaded to mark...
And another incident during startup of cunkserver, scanning HDD with 20 million chunks and partially paralyzing client/application for 15 minutes. Every time it is completely unexpected as scanning chunkserver have...
I've eliminated the problem so I won't be able to reproduce any more... The root cause as I understood it was chunkserver's CPU hog during iteration over great number of...
Both latest release 3.0.109 of course.