Open qiuqiu-007 opened 3 weeks ago
And when I stop the running backup task, here the rate increased shardly, such as
Hi @qiuqiu-007, thanks for reporting!
In my opinion, this rate limit is per node, or is it not?
It should work per node.
Could you send metric data used for evaluating those prom queries? It would also be good to see SM logs from the time of running the backup.
Log is missed. But, it looks that the uploaded bytes is increased then decreased, is that means the backup file missed ? or others?
I've found it was restarted by OOM, which caused backup file not found. By the way, is there any cpu memory set or limit suggestion? @Michal-Leszczynski
I've found it was restarted by OOM, which caused backup file not found. By the way, is there any cpu memory set or limit suggestion? @Michal-Leszczynski
which version are you using? Are you running the agent under some cgroup (that's a valid way to limit both its memory and CPU usage).
My Scylla manager version is 3.0, and scylla manager agent is also in running. OOM restart occured when I backup more than 600GB file bytes with command sctool backup -c major2 -L 's3:test-backup' -K 'alternator_test-new-v4' --rate-limit 50
, and my cpu is set 5 ,memory is 5G.
@qiuqiu-007 the newest SM is 3.2 which contains many changes since 3.0, so please update your SM version and see if the problems persist.
I execute the following command
sctool backup -c major1 -L 's3:scylla-backup' -K 'test-v4' --rate-limit 50
but in monitoring it is much lager than 50
In my opinion, this rate limit is per node, or is it not?