onlyjob
Yes, @borkd I have the same issues. Also spikes can be due to bugs like #273. The problem with vertically unaligned charts is that similar hardware don't look similar and...
The flaw in the current charts design is obvious on "cpu usage" graph where none of the graphs have 100% at the highest point. Because of different scale of each...
MooseFS removes chunks from disks marked for removal just not as aggressively as on normal disks. I observed that many times in CGI where counter or chunks on MFR disk...
A lot of thing is constantly going on in the cluster. CI jobs are running and large number of deletions are likely to happen at any time. There are 30_000...
Interesting. So I might have restarted chunkserver before realising that HDD marked for removal is already mounted read-only?
Empirical observation suggests that _deletetion_ is not happening anywhere in the cluster as long as one chunkserver is scanning at least one HDD.
Thank you. > Maybe we should redefine load and take into account also number of busy workers? That might be a good idea. > I don't think that CS is...
I've made some changes and for few days chunkserver is running with `WORKERS_MAX=1000` on _f2fs_ yet it remains in OVERLOADED state most of the time...
Yes in "Resources" tab, as I've said in the very first message... Just one device in the `mfshdd.cfg` so there is no internal rebalancing...
Thanks, @acid-maker. Your explanation make sense and it looks like I might be facing this very issue. New label `T` has been made specifically for this chunkserver as it is...