onlyjob

About 5 hours later chunkserver re-connected with master: long loop detected (14977.132197s)

> The 3% in 23 hours - I cannot explain that, it never happened to us. There must be some additional factor, I already mentioned this factor here. The problem...

And more troubles: remember that HDD with 5 million duplicates, which were waiting to be deleted for a week then being deleted for hours while chunkserver was offline? Just now...

Thanks for all the corrections and fixes in 3.0.110. Some things are definitely better but one thing is not (and perhaps a little worse) -- scanning of HDD full of...

Also I strongly suspect that duplicate chunks postponed for removal are not used in case when all active chunks are lost due to CRC check or "bad header" errors. It...

In regards to slow scanning, long loops and chunkserver's high CPU utilisation, it is definitely due to number of chunks per chunkserver: I've created a new chunkserver on the same...

Duplicates handling is still somewhat flawed. On chunkserver with only SSD disks and about a million chunks, presence of 30_000 duplicates throws frequent "long loop detected" and "replicator: connection lost"...

> Most admins will likely know how paranoid they need to be tracking access times. Or not. I provide storage space to team/developers and I have little knowledge of how...

IMHO "exit mode" is an anti-feature. How can I disable it? It already amplified duration of downtime on some occasions. It might make sense to activate "exit mode" on read-only...

Being able to switch to log-scale might be useful. However I would defend the original proposal to allow equalising charts (not necessarily by default but perhaps as an option such...