onlyjob

They were rendered on chart as replications, indistinguishable from normal (successful ones). It looked as abnormal spike on chart due to its hight. Exactly like in #452 where failed attempts...

In the same colour as the usual operations but much higher. On one occasion logs may have produced large number of "chunk already exist" messages or something like this, as...

 Here is an example of internal rebalance graph with bogus (attempted) operations visualised - see that abnormal spike around 22:15? A disk was marked as ` /mnt/03/ext4/ (can't find...

No, no it is all wrong... It is fair that manual action is required in case of master failure. For example primary master is down and we do no know...

Yes that would be useful. Thanks. I had replication running for over two weeks and statistics generated per chunk loop is definitely useful in a long run.

Maybe high load develops over time? Basically chunkserver became unresponsive causing logs flood on other chunkservers: replicator,get # of blocks: got status: Operation not completed from (192.168.0.250:72EF) Could it be...

Yes, there are some snapshots but only recent ones.

Very interesting. Thanks for explanation. It make sense. Current behaviour could be pretty bad because as soon as I make a snapshot (and change a storage class of the snapshot...

I tried to pin storage class `2` to "good" chunkservers but regardless, whenever I stop any archival chunkserver there are undergoal chunks in storage class `2`. So currently, unfortunately, some...

After thinking about the right way to solve this problem I came to conclusion that the only correct way of handling storage class of a snapshot would be to obey...