[jira] [Updated] (CASSANDRA-19485) if max_hints_size_per_host < max_hints_file_size then it will write hints after max_hints_size_per_host is reached
[ https://issues.apache.org/jira/browse/CASSANDRA-19485?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Stefan Miklosovic updated CASSANDRA-19485: -- Fix Version/s: (was: 5.x) (was: 4.1.x) (was: 5.0.x) > if max_hints_size_per_host < max_hints_file_size then it will write hints > after max_hints_size_per_host is reached > -- > > Key: CASSANDRA-19485 > URL: https://issues.apache.org/jira/browse/CASSANDRA-19485 > Project: Cassandra > Issue Type: Bug > Components: Consistency/Hints >Reporter: Stefan Miklosovic >Priority: Normal > > there is one problem in the current solution being that if we have this config > {noformat} > max_hints_size_per_host: 2MiB > max_hints_file_size: 128MiB > {noformat} > basically, max size > size per host, then it will not stop it from writing > hints after 2MiB for a particular node, because HintsDescriptor is added > among dispatching ones after writer is closed, which happens after there is > 128MiB written (all logic in HintsStore), so it will not be included into > total sizes, it will be 0 until at least one hints file for that node is > written to disk. > I consider this to be the flaw of CASSANDRA-17142, however it is questionable > if this is serious enough to deal with in the first place, I don't think > somebody would set it up in practice to these values, normally one puts there > like max per host is few gigs so this problem is not so visible but it shows > in tests almost instantly and it is technically just wrong, regardless of the > probability this would happen in real ... > cc [~yifanc] -- This message was sent by Atlassian Jira (v8.20.10#820010) - To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org For additional commands, e-mail: commits-h...@cassandra.apache.org
[jira] [Updated] (CASSANDRA-19485) if max_hints_size_per_host < max_hints_file_size then it will write hints after max_hints_size_per_host is reached
[ https://issues.apache.org/jira/browse/CASSANDRA-19485?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Stefan Miklosovic updated CASSANDRA-19485: -- Resolution: Resolved Status: Resolved (was: Triage Needed) The fix for this is done as part of CASSANDRA-19477 > if max_hints_size_per_host < max_hints_file_size then it will write hints > after max_hints_size_per_host is reached > -- > > Key: CASSANDRA-19485 > URL: https://issues.apache.org/jira/browse/CASSANDRA-19485 > Project: Cassandra > Issue Type: Bug > Components: Consistency/Hints >Reporter: Stefan Miklosovic >Priority: Normal > Fix For: 4.1.x, 5.0.x, 5.x > > > there is one problem in the current solution being that if we have this config > {noformat} > max_hints_size_per_host: 2MiB > max_hints_file_size: 128MiB > {noformat} > basically, max size > size per host, then it will not stop it from writing > hints after 2MiB for a particular node, because HintsDescriptor is added > among dispatching ones after writer is closed, which happens after there is > 128MiB written (all logic in HintsStore), so it will not be included into > total sizes, it will be 0 until at least one hints file for that node is > written to disk. > I consider this to be the flaw of CASSANDRA-17142, however it is questionable > if this is serious enough to deal with in the first place, I don't think > somebody would set it up in practice to these values, normally one puts there > like max per host is few gigs so this problem is not so visible but it shows > in tests almost instantly and it is technically just wrong, regardless of the > probability this would happen in real ... > cc [~yifanc] -- This message was sent by Atlassian Jira (v8.20.10#820010) - To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org For additional commands, e-mail: commits-h...@cassandra.apache.org
[jira] [Updated] (CASSANDRA-19485) if max_hints_size_per_host < max_hints_file_size then it will write hints after max_hints_size_per_host is reached
[ https://issues.apache.org/jira/browse/CASSANDRA-19485?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Stefan Miklosovic updated CASSANDRA-19485: -- Fix Version/s: 4.1.x 5.0.x 5.x > if max_hints_size_per_host < max_hints_file_size then it will write hints > after max_hints_size_per_host is reached > -- > > Key: CASSANDRA-19485 > URL: https://issues.apache.org/jira/browse/CASSANDRA-19485 > Project: Cassandra > Issue Type: Bug >Reporter: Stefan Miklosovic >Priority: Normal > Fix For: 4.1.x, 5.0.x, 5.x > > > there is one problem in the current solution being that if we have this config > {noformat} > max_hints_size_per_host: 2MiB > max_hints_file_size: 128MiB > {noformat} > basically, max size > size per host, then it will not stop it from writing > hints after 2MiB for a particular node, because HintsDescriptor is added > among dispatching ones after writer is closed, which happens after there is > 128MiB written (all logic in HintsStore), so it will not be included into > total sizes, it will be 0 until at least one hints file for that node is > written to disk. > I consider this to be the flaw of CASSANDRA-17142, however it is questionable > if this is serious enough to deal with in the first place, I don't think > somebody would set it up in practice to these values, normally one puts there > like max per host is few gigs so this problem is not so visible but it shows > in tests almost instantly and it is technically just wrong, regardless of the > probability this would happen in real ... > cc [~yifanc] -- This message was sent by Atlassian Jira (v8.20.10#820010) - To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org For additional commands, e-mail: commits-h...@cassandra.apache.org
[jira] [Updated] (CASSANDRA-19485) if max_hints_size_per_host < max_hints_file_size then it will write hints after max_hints_size_per_host is reached
[ https://issues.apache.org/jira/browse/CASSANDRA-19485?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Stefan Miklosovic updated CASSANDRA-19485: -- Component/s: Consistency/Hints > if max_hints_size_per_host < max_hints_file_size then it will write hints > after max_hints_size_per_host is reached > -- > > Key: CASSANDRA-19485 > URL: https://issues.apache.org/jira/browse/CASSANDRA-19485 > Project: Cassandra > Issue Type: Bug > Components: Consistency/Hints >Reporter: Stefan Miklosovic >Priority: Normal > Fix For: 4.1.x, 5.0.x, 5.x > > > there is one problem in the current solution being that if we have this config > {noformat} > max_hints_size_per_host: 2MiB > max_hints_file_size: 128MiB > {noformat} > basically, max size > size per host, then it will not stop it from writing > hints after 2MiB for a particular node, because HintsDescriptor is added > among dispatching ones after writer is closed, which happens after there is > 128MiB written (all logic in HintsStore), so it will not be included into > total sizes, it will be 0 until at least one hints file for that node is > written to disk. > I consider this to be the flaw of CASSANDRA-17142, however it is questionable > if this is serious enough to deal with in the first place, I don't think > somebody would set it up in practice to these values, normally one puts there > like max per host is few gigs so this problem is not so visible but it shows > in tests almost instantly and it is technically just wrong, regardless of the > probability this would happen in real ... > cc [~yifanc] -- This message was sent by Atlassian Jira (v8.20.10#820010) - To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org For additional commands, e-mail: commits-h...@cassandra.apache.org
[jira] [Updated] (CASSANDRA-19485) if max_hints_size_per_host < max_hints_file_size then it will write hints after max_hints_size_per_host is reached
[ https://issues.apache.org/jira/browse/CASSANDRA-19485?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Stefan Miklosovic updated CASSANDRA-19485: -- Description: there is one problem in the current solution being that if we have this config {noformat} max_hints_size_per_host: 2MiB max_hints_file_size: 128MiB {noformat} basically, max size > size per host, then it will not stop it from writing hints after 2MiB for a particular node, because HintsDescriptor is added among dispatching ones after writer is closed, which happens after there is 128MiB written (all logic in HintsStore), so it will not be included into total sizes, it will be 0 until at least one hints file for that node is written to disk. I consider this to be the flaw of CASSANDRA-17142, however it is questionable if this is serious enough to deal with in the first place, I don't think somebody would set it up in practice to these values, normally one puts there like max per host is few gigs so this problem is not so visible but it shows in tests almost instantly and it is technically just wrong, regardless of the probability this would happen in real ... cc [~yifanc] > if max_hints_size_per_host < max_hints_file_size then it will write hints > after max_hints_size_per_host is reached > -- > > Key: CASSANDRA-19485 > URL: https://issues.apache.org/jira/browse/CASSANDRA-19485 > Project: Cassandra > Issue Type: Bug >Reporter: Stefan Miklosovic >Priority: Normal > > there is one problem in the current solution being that if we have this config > {noformat} > max_hints_size_per_host: 2MiB > max_hints_file_size: 128MiB > {noformat} > basically, max size > size per host, then it will not stop it from writing > hints after 2MiB for a particular node, because HintsDescriptor is added > among dispatching ones after writer is closed, which happens after there is > 128MiB written (all logic in HintsStore), so it will not be included into > total sizes, it will be 0 until at least one hints file for that node is > written to disk. > I consider this to be the flaw of CASSANDRA-17142, however it is questionable > if this is serious enough to deal with in the first place, I don't think > somebody would set it up in practice to these values, normally one puts there > like max per host is few gigs so this problem is not so visible but it shows > in tests almost instantly and it is technically just wrong, regardless of the > probability this would happen in real ... > cc [~yifanc] -- This message was sent by Atlassian Jira (v8.20.10#820010) - To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org For additional commands, e-mail: commits-h...@cassandra.apache.org
[jira] [Updated] (CASSANDRA-19485) if max_hints_size_per_host < max_hints_file_size then it will write hints after max_hints_size_per_host is reached
[ https://issues.apache.org/jira/browse/CASSANDRA-19485?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Stefan Miklosovic updated CASSANDRA-19485: -- Summary: if max_hints_size_per_host < max_hints_file_size then it will write hints after max_hints_size_per_host is reached (was: if max_hints_size_per_host < max_hints_file_size then it will write hints after max_hints_file_size is reached) > if max_hints_size_per_host < max_hints_file_size then it will write hints > after max_hints_size_per_host is reached > -- > > Key: CASSANDRA-19485 > URL: https://issues.apache.org/jira/browse/CASSANDRA-19485 > Project: Cassandra > Issue Type: Bug >Reporter: Stefan Miklosovic >Priority: Normal > -- This message was sent by Atlassian Jira (v8.20.10#820010) - To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org For additional commands, e-mail: commits-h...@cassandra.apache.org