RE: Apt repositories
If you don't want your APT-sourced packages to upgrade automatically, I suggest pinning the package. The apt_preferences(5) man page tells you how to do this. The gist is to add the following lines: Package: cassandra Pin: version 0.6.13 Pin-Priority: 1100 (setting the version to the one you want to install, obviously) to a preferences file sourced by apt. On Ubuntu, just place the above 3 lines in the file /etc/apt/preferences.d/cassandra and you should be set. No matter what happens with the remote APT repository or how you run `apt-get upgrade`, your system will always use the version you specified in the preferences file. Greg -Original Message- From: David Strauss [mailto:da...@davidstrauss.net] Sent: Saturday, April 23, 2011 4:49 PM To: user@cassandra.apache.org Subject: Apt repositories I just noticed that, following the Cassandra 0.8 beta release, the Apt repository is encouraging servers in my clusters to upgrade. Beta releases should probably be on different channels (or named differently) than stable ones. Better yet would be naming the packages based on the major release in order to prevent an inadvertent upgrade, even once the next release stabilizes. For example, having cassandra-0.7 and cassandra-0.8 would be great, with installation of the latter replacing any cassandra-0.7 package. This is common with PHP and MySQL packages where it's not entirely safe to inadvertently do a major upgrade. Thanks, David
RE: Nodes frozen in GC
I do believe there is a fundamental issue with compactions allocating too much memory and incurring too many garbage collections (at least with 0.6.12). On nearly every Cassandra node I operate, garbage collections simply get out of control during compactions of any reasonably sized CF (1GB). I can reproduce it on CF's with many wider rows (1000's of columns) consisting of smaller columns (10's-100's of bytes) and CF's with thinner rows (20 columns) with larger columns (10's MBs) and everything in between. From the GC logs, I can infer that Cassandra is allocating upwards of 4GB/s. I once gave the JVM 30GB of heap and saw it run through the entire heap in a few seconds while doing a compaction! It would continuously blow through the heap, incur a stop-the-world collection, and repeat. Meanwhile, the listed compacted bytes from the JMX interface was never increasing and the tmp sstable wasn't growing in size. My current/relevant JVM args are as follows (running on Sun 1.6.0.24 w/ JNA 3.2.7): -Xms9G -Xmx9G -Xmn256M -XX:+PrintGCDetails -XX:+PrintGCTimeStamps -XX:+PrintClassHistogram -XX:+PrintTenuringDistribution -Xloggc:/var/log/cassandra/gc.log -XX:+UseParNewGC -XX:+UseConcMarkSweepGC -XX:+CMSParallelRemarkEnabled -XX:SurvivorRatio=8 -XX:MaxTenuringThreshold=3 -XX:CMSInitiatingOccupancyFraction=40 -XX:+HeapDumpOnOutOfMemoryError -XX:+UseCMSInitiatingOccupancyOnly -XX:CMSFullGCsBeforeCompaction=1 -XX:ParallelGCThreads=6 I've tweaked with nearly every setting imaginable (http://www.md.pp.ru/~eu/jdk6options.html is a great resource, BTW) and can't control the problem. No matter what I do, nothing can solve the problem of Cassandra allocating objects faster than the GC can clean them. And, when we're talking about 1GB/s of allocations, I don't think you can blame GC for not keeping up. Since there is no way to prevent these frequent stop-the-world collections, we get frequent client timeouts and an occasional unavailable response if we're unfortunate to have a couple of nodes compacting large CFs at the same time (which happens more than I'd like). For the past two weeks, we had N=replication factor adjacent nodes in our cluster that failed to perform their daily major compaction on a particular column family. All N would spew GCInspector logs and the GC logs revealed heavy memory allocation rate. The only resolution was to restart Cassandra to abort the compaction. I isolated one node from network connectivity and restarted it in a cluster of 1 with no caching, memtables, or any operations. Under these ideal compacting conditions, I still ran into issues. I experimented with extremely large young generations (up to 10GB), very low CMSInitiatingOccupancyFraction, etc, but Cassandra would always allocate faster than JVM could collect, eventually leading to stop-the-world. Recently, we rolled out a change to the application accessing the cluster which effectively resaved every column in every row. When this was mostly done, our daily major compaction for the trouble CF that refused to compact for two weeks, suddenly completed! Most interesting. (Although, it still went through memory to no end.) One of my observations is that memory allocations during compaction seems to be mostly short-lived objects. The young generation is almost never promoting objects to the tenured generation (we changed our MaxTenuringThreshold to 3, from Cassandra's default of 1 to discourage early promotion- a default of 1 seems rather silly to me). However, when the young generation is being collected (which happens VERY often during compactions b/c allocation rate is so high), objects are allocated directly into the tenured generation. Even with relatively short ParNew collections (often 0.05s, almost always 0.1s wall time), these tenured allocations quickly accumulate, initiating CMS and eventually stop-the-world. Anyway, not sure how much additional writing is going to help resolve this issue. I have gobs of GC logs and supplementary metrics data to back up my claims if those will help. But, I have a feeling that if you just create a CF of a few GB and incur a compaction with the JVM under a profiler, it will be pretty easy to identify the culprit. I've started down this path and will let you know if I find anything. But, I'm no Java expert and am quite busy with other tasks, so don't expect anything useful from me anytime soon. I hope this information helps. If you need anything else, just ask, and I'll see what I can do. Gregory Szorc gregory.sz...@xobni.com -Original Message- From: sc...@scode.org [mailto:sc...@scode.org] On Behalf Of Peter Schuller Sent: Thursday, March 10, 2011 10:36 AM To: ruslan usifov Cc: user@cassandra.apache.org Subject: Re: Nodes frozen in GC I think it would be very useful to get to the bottom of this but without further details (like the asked for GC logs) I'm not sure what to do/suggest. It's clear that a single CF
RE: Request For 0.6.12 Release
Aaron, Thank you very much for initiating the voting process. I'm looking forward to running this release. Was there any discussion around improving the communication of known issues with releases? Gregory From: Aaron Morton [mailto:aa...@thelastpickle.com] Sent: Thursday, February 17, 2011 4:21 PM To: user@cassandra.apache.org Subject: Re: Request For 0.6.12 Release Gregory, There is a vote going on for 0.6.12 now http://www.mail-archive.com/dev@cassandra.apache.org/msg01808.html If you have time grab the bin and give it a test http://people.apache.org/~eevans Aaron On 16 Feb, 2011,at 09:21 PM, Aaron Morton aa...@thelastpickle.commailto:aa...@thelastpickle.com wrote: Have checked it's all in the 0.6 branch and asked the devs for a 0.6.12 release. Will let you know how it goes. cheers Aaron On 16 Feb, 2011,at 08:38 AM, Aaron Morton aa...@thelastpickle.commailto:aa...@thelastpickle.com wrote: I worked on that ticket, will try to chase it up. Aaron On 15/02/2011, at 2:01 PM, Gregory Szorc gregory.sz...@gmail.commailto:gregory.sz...@gmail.com wrote: The latest official 0.6.x releases, 0.6.10 and 0.6.11, have a very serious bug/regression when performing some quorum reads (CASSANDRA-2081), which is fixed in the head of the 0.6 branch If there aren't any plans to cut 0.6.12 any time soon, as an end user, I request that an official and blessed release of 0.6.x be made ASAP. On a related note, I am frustrated that such a serious issue has lingered in the latest oldstable release. I would have liked to see one or more of the following: 1) The issue documented prominently on the apache.orghttp://apache.org web site and inside the download archive so end users would know they are downloading and running known-broken software 2) The 0.6.10 and 0.6.11 builds pulled after identification of the issue 3) A 0.6.12 release cut immediately (with reasonable time for testing, of course) to address the issue I understand that releases may not always be as stable as we all desire. But, I hope that when future bugs affecting the bread and butter properties of a distributed storage engine surface (especially when they are regressions) that the official project response (preferably via mailing list and the web site) is swift and maximizes the potential for data integrity and availability. If there is anything I can do to help the process, I'd gladly give some of my time to help the overall community. Gregory Szorc gregory.sz...@gmail.commailto:gregory.sz...@gmail.com
Request For 0.6.12 Release
The latest official 0.6.x releases, 0.6.10 and 0.6.11, have a very serious bug/regression when performing some quorum reads (CASSANDRA-2081), which is fixed in the head of the 0.6 branch. If there aren't any plans to cut 0.6.12 any time soon, as an end user, I request that an official and blessed release of 0.6.x be made ASAP. On a related note, I am frustrated that such a serious issue has lingered in the latest oldstable release. I would have liked to see one or more of the following: 1) The issue documented prominently on the apache.org web site and inside the download archive so end users would know they are downloading and running known-broken software 2) The 0.6.10 and 0.6.11 builds pulled after identification of the issue 3) A 0.6.12 release cut immediately (with reasonable time for testing, of course) to address the issue I understand that releases may not always be as stable as we all desire. But, I hope that when future bugs affecting the bread and butter properties of a distributed storage engine surface (especially when they are regressions) that the official project response (preferably via mailing list and the web site) is swift and maximizes the potential for data integrity and availability. If there is anything I can do to help the process, I'd gladly give some of my time to help the overall community. Gregory Szorc gregory.sz...@gmail.com