Roy Dahan comments

Results 453 comments of


                                            Roy Dahan

'Cannot perform truncate, some hosts are down' with hosts that have left the cluster

You copy this job "[Jenkins job URL](https://jenkins.scylladb.com/job/scylla-master/job/longevity/job/longevity-twcs-48h-test/25/)" as a reproducer job in Jenkins and run it. To find the minimal reproducer you need to suggest what could be the cause....

reader_concurrency_semaphore time outs on various Scylla nodes running on centos7 during the rolling upgrade

@slivne / @DoronArazii, this one also need to be looked at. relevant for 5.0 & 2022.1

reader_concurrency_semaphore time outs on various Scylla nodes running on centos7 during the rolling upgrade

> how are we setting up the n1-highmem-8 machines - do they have a local ssd at all Yes, we add and setup Scylla to use local ssd.

reader_concurrency_semaphore time outs on various Scylla nodes running on centos7 during the rolling upgrade

@vponomaryov can you please check if the problem comes from our side?

Memtable flush aborts the node if it fails due to out-of storage space

@xemul do you need our help with this?

scylla_util.py: on sysconfig_parser, don't use double quote when it's possible

IIRC we recommend to have it as disabled and I think that the issue here was that it wasn’t disabled.

reader_concurrency_semaphore time outs on various Scylla nodes running on centos7 during the rolling upgrade

We can close. The problem is that it's hard for us to identify if timeouts we see are new or similar to older issues.

Failed to obtain stats (seastar::metrics::double_registration (registering metrics twice for metrics: storage_proxy_coordinator_background_replica_writes_failed_remote_node)), fall-back to dummy

@eliransin / @elcallio Is this something someone is going to fix? If not, we're going to set SCT to ignore this error.

Failed to obtain stats (seastar::metrics::double_registration (registering metrics twice for metrics: storage_proxy_coordinator_background_replica_writes_failed_remote_node)), fall-back to dummy

I just bump into it with 5.2.0.rc1, that's why I commented here. If it's not part of 5.2, I can re-run in master.

Crash/assert in seastar::aio_general_context::flush(): Assertion `clock::now() < retry_until' failed.

@avikivity / @bhalevy would it be ok to workaround this configuration if we one set --developer-mode? It's the most common use case for people who uses docker and it's blocking...