Israel Fruchter
Israel Fruchter
> Just to avoid catching all places we assumed it's centos. > But if it should be simple with docker backend it should be easier. > Only thing left is...
Anyhow lets wait for more incidents of this, before deciding we should more in any of the directions
> Anyhow lets wait for more incidents of this, before deciding we should more in any of the directions The only downside of waiting, is that is might affect next...
I'm guessing kind of related: docker service, coredumped: ``` Mar 27 22:50:25 gemini-with-nemesis-3h-normal-maste-loader-node-59828a1f-1 systemd-coredump[24575]: Failed to create coredump file /var/lib/systemd/coredump/.#core.dockerd.0.f407a0f9fb9244698afc85d58475c463.1364.167995742400000010c8b3daaeed3160: No such file or directory Mar 27 22:50:25 gemini-with-nemesis-3h-normal-maste-loader-node-59828a1f-1 systemd-coredump[24575]:...
happened again in next: https://jenkins.scylladb.com/job/scylla-master/job/next/7185/testReport/junit/update_cluster_layout_tests/TestUpdateClusterLayout/Build___x86___dtest___test_add_node_with_large_partition2/
it's happening once in a while in those tests for the last 6 months: https://70f106c98484448dbc4705050eb3f7e9.us-east-1.aws.found.io:9243/goto/0ec7dc70-c4c3-11ee-81c7-3986d18dafd5 hinted handoff is supposed to be disabled for those tests, what else case cause overload...
two other places in the code are: - https://github.com/scylladb/scylladb/blob/bd3ed168ab79c9ea32a989b1815cf380637e8a0a/transport/server.cc#L680 - https://github.com/scylladb/scylladb/blob/bd3ed168ab79c9ea32a989b1815cf380637e8a0a/transport/server.cc#L714
> > @bhalevy any idea? 100 * 1k = 100k data in flight. This is nothing! > > @avikivity I think this is the wrong calculation - The insert c1_cn...
in all of dtest we have 64 calls to `execute_concurrent_with_args` only 17 of them are setting concurrency to something other then the default.
> I briefly checked what the `ConnectionBusy` exception is about. It is actually generated by the python driver, apparently when it cannot write to the socket. > > It is...