[jira] [Commented] (CASSANDRA-5306) Improve Dsnitch Severity
[ https://issues.apache.org/jira/browse/CASSANDRA-5306?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13611602#comment-13611602 ] Jonathan Ellis commented on CASSANDRA-5306: --- I'd be happy with just iowait, to be honest, which can be easily read from /proc. That would be a big improvement on the existing heuristic; Windows can fall back to compaction throughput. The current heuristic of "percentage of my space that I'm currently compacting" is only really useful for distinguishing "compacting" from "not compacting;" whether it will give a useful comparison between two different nodes otherwise is essentially random. > Improve Dsnitch Severity > > > Key: CASSANDRA-5306 > URL: https://issues.apache.org/jira/browse/CASSANDRA-5306 > Project: Cassandra > Issue Type: Improvement >Reporter: Vijay >Assignee: Vijay >Priority: Minor > Fix For: 1.2.4 > > > This ticket is to continue the discussion in CASSANDRA-5255. > Currently Dsnitch reports the Severity by calculating the amount of data > compacting relative to load of the node. > It will be nice to report Severity as a factor of load average, N/W > Throughput and IO Wait instead. > http://www.hyperic.com/products/sigar seem to have it (But personally i have > not used it). -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-5306) Improve Dsnitch Severity
[ https://issues.apache.org/jira/browse/CASSANDRA-5306?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13611603#comment-13611603 ] Jonathan Ellis commented on CASSANDRA-5306: --- (OCD: I would also be in favor of naming it something like BackgroundActivity instead of Severity...) > Improve Dsnitch Severity > > > Key: CASSANDRA-5306 > URL: https://issues.apache.org/jira/browse/CASSANDRA-5306 > Project: Cassandra > Issue Type: Improvement >Reporter: Vijay >Assignee: Vijay >Priority: Minor > Fix For: 1.2.4 > > > This ticket is to continue the discussion in CASSANDRA-5255. > Currently Dsnitch reports the Severity by calculating the amount of data > compacting relative to load of the node. > It will be nice to report Severity as a factor of load average, N/W > Throughput and IO Wait instead. > http://www.hyperic.com/products/sigar seem to have it (But personally i have > not used it). -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-5306) Improve Dsnitch Severity
[ https://issues.apache.org/jira/browse/CASSANDRA-5306?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13612730#comment-13612730 ] Jonathan Ellis commented on CASSANDRA-5306: --- iowait code looks good. Still need to switch compaction info to throughput instead of load / size. Other nits: - Catch should log at info (warn?) that it couldn't open the stats file. (Better: be smart and only log warn if Linux.) - Methods that are only used internally by BAM should be private - empty catch should log ioexception at debug, just in case > Improve Dsnitch Severity > > > Key: CASSANDRA-5306 > URL: https://issues.apache.org/jira/browse/CASSANDRA-5306 > Project: Cassandra > Issue Type: Improvement >Reporter: Vijay >Assignee: Vijay >Priority: Minor > Fix For: 1.2.4 > > Attachments: 0001-5306.patch > > > This ticket is to continue the discussion in CASSANDRA-5255. > Currently Dsnitch reports the Severity by calculating the amount of data > compacting relative to load of the node. > It will be nice to report Severity as a factor of load average, N/W > Throughput and IO Wait instead. > http://www.hyperic.com/products/sigar seem to have it (But personally i have > not used it). -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-5306) Improve Dsnitch Severity
[ https://issues.apache.org/jira/browse/CASSANDRA-5306?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13613073#comment-13613073 ] Vijay commented on CASSANDRA-5306: -- {quote} Still need to switch compaction info to throughput instead of load / size. {quote} Are we talking about (user defined throttle / compacting file sizes)? if yes then we will be penalizing instances which can do more on disk... IMHO: the bigger the compaction bigger the file cache misses and more memory pressure, hence we added the size/load in the first place. > Improve Dsnitch Severity > > > Key: CASSANDRA-5306 > URL: https://issues.apache.org/jira/browse/CASSANDRA-5306 > Project: Cassandra > Issue Type: Improvement >Reporter: Vijay >Assignee: Vijay >Priority: Minor > Fix For: 1.2.4 > > Attachments: 0001-5306.patch > > > This ticket is to continue the discussion in CASSANDRA-5255. > Currently Dsnitch reports the Severity by calculating the amount of data > compacting relative to load of the node. > It will be nice to report Severity as a factor of load average, N/W > Throughput and IO Wait instead. > http://www.hyperic.com/products/sigar seem to have it (But personally i have > not used it). -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-5306) Improve Dsnitch Severity
[ https://issues.apache.org/jira/browse/CASSANDRA-5306?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13622442#comment-13622442 ] Jonathan Ellis commented on CASSANDRA-5306: --- That makes sense. Did getSystemLoadAverage turn out to not be sufficiently realtime? > Improve Dsnitch Severity > > > Key: CASSANDRA-5306 > URL: https://issues.apache.org/jira/browse/CASSANDRA-5306 > Project: Cassandra > Issue Type: Improvement >Reporter: Vijay >Assignee: Vijay >Priority: Minor > Fix For: 1.2.4 > > Attachments: 0001-5306.patch > > > This ticket is to continue the discussion in CASSANDRA-5255. > Currently Dsnitch reports the Severity by calculating the amount of data > compacting relative to load of the node. > It will be nice to report Severity as a factor of load average, N/W > Throughput and IO Wait instead. > http://www.hyperic.com/products/sigar seem to have it (But personally i have > not used it). -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-5306) Improve Dsnitch Severity
[ https://issues.apache.org/jira/browse/CASSANDRA-5306?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13622518#comment-13622518 ] Vijay commented on CASSANDRA-5306: -- Hi Jonathan, yeah it wasn't that sensitive as much as IOWait, some moderate disk latency spikes disappear and wont show up... > Improve Dsnitch Severity > > > Key: CASSANDRA-5306 > URL: https://issues.apache.org/jira/browse/CASSANDRA-5306 > Project: Cassandra > Issue Type: Improvement >Reporter: Vijay >Assignee: Vijay >Priority: Minor > Fix For: 1.2.4 > > Attachments: 0001-5306.patch > > > This ticket is to continue the discussion in CASSANDRA-5255. > Currently Dsnitch reports the Severity by calculating the amount of data > compacting relative to load of the node. > It will be nice to report Severity as a factor of load average, N/W > Throughput and IO Wait instead. > http://www.hyperic.com/products/sigar seem to have it (But personally i have > not used it). -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-5306) Improve Dsnitch Severity
[ https://issues.apache.org/jira/browse/CASSANDRA-5306?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13644606#comment-13644606 ] Jonathan Ellis commented on CASSANDRA-5306: --- I guess "that makes sense" wasn't very clear, so +1. (Retargetting for 2.0 -- let's go ahead and rename it there. :) > Improve Dsnitch Severity > > > Key: CASSANDRA-5306 > URL: https://issues.apache.org/jira/browse/CASSANDRA-5306 > Project: Cassandra > Issue Type: Improvement >Reporter: Vijay >Assignee: Vijay >Priority: Minor > Fix For: 1.2.5 > > Attachments: 0001-5306.patch > > > This ticket is to continue the discussion in CASSANDRA-5255. > Currently Dsnitch reports the Severity by calculating the amount of data > compacting relative to load of the node. > It will be nice to report Severity as a factor of load average, N/W > Throughput and IO Wait instead. > http://www.hyperic.com/products/sigar seem to have it (But personally i have > not used it). -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-5306) Improve Dsnitch Severity
[ https://issues.apache.org/jira/browse/CASSANDRA-5306?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13649179#comment-13649179 ] Marcus Eriksson commented on CASSANDRA-5306: seeing intermittent failures in DynamicEndpointSnitchTest: [junit] Testcase: testSnitch(org.apache.cassandra.locator.DynamicEndpointSnitchTest): FAILED [junit] null [junit] junit.framework.AssertionFailedError [junit] at org.apache.cassandra.locator.DynamicEndpointSnitchTest.testSnitch(DynamicEndpointSnitchTest.java:61) [junit] [junit] [junit] Test org.apache.cassandra.locator.DynamicEndpointSnitchTest FAILED it seems to be related to the amount of iowait there is on the host running the test - high iowait means the list does not sort the expected way. > Improve Dsnitch Severity > > > Key: CASSANDRA-5306 > URL: https://issues.apache.org/jira/browse/CASSANDRA-5306 > Project: Cassandra > Issue Type: Improvement >Reporter: Vijay >Assignee: Vijay >Priority: Minor > Fix For: 2.0 > > Attachments: 0001-5306.patch > > > This ticket is to continue the discussion in CASSANDRA-5255. > Currently Dsnitch reports the Severity by calculating the amount of data > compacting relative to load of the node. > It will be nice to report Severity as a factor of load average, N/W > Throughput and IO Wait instead. > http://www.hyperic.com/products/sigar seem to have it (But personally i have > not used it). -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-5306) Improve Dsnitch Severity
[ https://issues.apache.org/jira/browse/CASSANDRA-5306?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13649488#comment-13649488 ] Vijay commented on CASSANDRA-5306: -- Marcus, I am not able to reproduce, can you give the following a try? in DEST from {code} InetAddress host1 = InetAddress.getByName("127.0.0.1"); {code} to {code} InetAddress host1 = InetAddress.getByName("127.0.0.4"); {code} > Improve Dsnitch Severity > > > Key: CASSANDRA-5306 > URL: https://issues.apache.org/jira/browse/CASSANDRA-5306 > Project: Cassandra > Issue Type: Improvement >Reporter: Vijay >Assignee: Vijay >Priority: Minor > Fix For: 2.0 > > Attachments: 0001-5306.patch > > > This ticket is to continue the discussion in CASSANDRA-5255. > Currently Dsnitch reports the Severity by calculating the amount of data > compacting relative to load of the node. > It will be nice to report Severity as a factor of load average, N/W > Throughput and IO Wait instead. > http://www.hyperic.com/products/sigar seem to have it (But personally i have > not used it). -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-5306) Improve Dsnitch Severity
[ https://issues.apache.org/jira/browse/CASSANDRA-5306?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13649522#comment-13649522 ] Marcus Eriksson commented on CASSANDRA-5306: I reproduced it by hard coding BackgroundActivityMonitor#getIOWait() to return 100.0; and yes, not using 127.0.0.1 fixes it > Improve Dsnitch Severity > > > Key: CASSANDRA-5306 > URL: https://issues.apache.org/jira/browse/CASSANDRA-5306 > Project: Cassandra > Issue Type: Improvement >Reporter: Vijay >Assignee: Vijay >Priority: Minor > Fix For: 2.0 > > Attachments: 0001-5306.patch > > > This ticket is to continue the discussion in CASSANDRA-5255. > Currently Dsnitch reports the Severity by calculating the amount of data > compacting relative to load of the node. > It will be nice to report Severity as a factor of load average, N/W > Throughput and IO Wait instead. > http://www.hyperic.com/products/sigar seem to have it (But personally i have > not used it). -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-5306) Improve Dsnitch Severity
[ https://issues.apache.org/jira/browse/CASSANDRA-5306?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13649532#comment-13649532 ] Vijay commented on CASSANDRA-5306: -- Committed the fix with 5633503d579b5fd67282b7e4d1f252489acb9e8, Thanks! > Improve Dsnitch Severity > > > Key: CASSANDRA-5306 > URL: https://issues.apache.org/jira/browse/CASSANDRA-5306 > Project: Cassandra > Issue Type: Improvement >Reporter: Vijay >Assignee: Vijay >Priority: Minor > Fix For: 2.0 > > Attachments: 0001-5306.patch > > > This ticket is to continue the discussion in CASSANDRA-5255. > Currently Dsnitch reports the Severity by calculating the amount of data > compacting relative to load of the node. > It will be nice to report Severity as a factor of load average, N/W > Throughput and IO Wait instead. > http://www.hyperic.com/products/sigar seem to have it (But personally i have > not used it). -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira