I deleted ALL data and reset the nodes from scratch. There are no more large rows in there. 8-9megs MAX across all nodes. This appears to be a new problem. I restarted the node in question and it seems to be running fine, but I had to run repair on it as it appears to be missing a lot of data.
On Fri, Aug 20, 2010 at 7:51 PM, Edward Capriolo <edlinuxg...@gmail.com>wrote: > On Fri, Aug 20, 2010 at 1:17 PM, Wayne <wav...@gmail.com> wrote: > > I turned off the creation of the secondary indexes which had the large > rows > > and all seemed good. Thank you for the help. I was getting > > 60k+/writes/second on the 6 node cluster. > > > > Unfortunately again three hours later a node went down. I can not even > look > > at the logs when it started since they are gone/recycled due to millions > of > > message deserialization messages. What are these? The node had 12,098,067 > > pending message-deserializer-pool entries in tpstats. The node was up > > according to some nodes and down according to others which made it > flapping > > and still trying to take requests. What is the log warning message > > deserialization task dropped message? Why would a node have 12 million of > > these? > > > > WARN [MESSAGE-DESERIALIZER-POOL:1] 2010-08-20 16:57:02,602 > > MessageDeserializationTask.java (line 47) dropping message (1,078,378ms > past > > timeout) > > WARN [MESSAGE-DESERIALIZER-POOL:1] 2010-08-20 16:57:02,602 > > MessageDeserializationTask.java (line 47) dropping message (1,078,378ms > past > > timeout) > > > > I do not think this is a large row problem any more. All nodes show a max > > row size around 8-9 megs. > > > > I looked at the munin charts and the disk IO seems to have spiked along > with > > compaction. Could compaction kicking in cause this? I have added the 3 > JVM > > settings to make compaction a lower priority. Did this help cause this to > > happen by slowing down and building up compaction on a heavily loaded > > system? > > > > Thanks in advance for any help someone can provide. > > > > > > On Fri, Aug 20, 2010 at 8:34 AM, Wayne <wav...@gmail.com> wrote: > >> > >> The NullPointerException does not crash the node. It only makes it > flap/go > >> down a for short period and then it comes back up. I do not see anything > >> abnormal in the system log, only that single error in the cassandra.log. > >> > >> > >> On Thu, Aug 19, 2010 at 11:42 PM, Peter Schuller > >> <peter.schul...@infidyne.com> wrote: > >>> > >>> > What is my "live set"? > >>> > >>> Sorry; that meant the "set of data acually live (i.e., not garbage) in > >>> the heap". In other words, the amount of memory truly "used". > >>> > >>> > Is the system CPU bound given the few statements > >>> > below? This is from running 4 concurrent processes against the > >>> > node...do I > >>> > need to throttle back the concurrent read/writers? > >>> > > >>> > I do all reads/writes as Quorum. (Replication factor of 3). > >>> > >>> With quorom and 0.6.4 I don't think unthrottled writes are expected to > >>> cause a problem. > >>> > >>> > The memtable threshold is the default of 256. > >>> > > >>> > All caching is turned off. > >>> > > >>> > The database is pretty small, maybe a few million keys (2-3) in 4 > CFs. > >>> > The > >>> > key size is pretty small. Some of the rows are pretty fat though > >>> > (fatter > >>> > than I thought). I am saving secondary indexes in separate CFs and > >>> > those are > >>> > the large rows that I think might be part of the problem. I will > >>> > restart > >>> > testing turning these off and see if I see any difference. > >>> > > >>> > Would an extra fat row explain repeated OOM crashes in a row? I have > >>> > finally > >>> > got the system to stabilize relatively and I even ran compaction on > the > >>> > bad > >>> > node without a problem (still no row size stats). > >>> > >>> Based on what you've said so far, the large rows are the only thing I > >>> would suspect may be the cause. With the amount of data and keys you > >>> say you have, you should definitely not be having memory issues with > >>> an 8 gig heap as a direct result of the data size/key count. A few > >>> million keys is not a lot at all; I still claim you should be able to > >>> handle hundreds of millions at least, from the perspective of bloom > >>> filters and such. > >>> > >>> So your plan to try it without these large rows is probably a good > >>> idea unless some else has a better idea. > >>> > >>> You may want to consider trying 0.7 betas too since it has removed the > >>> limitation with respect to large rows, assuming you do in fact want > >>> these large rows (see the CassandraLimitations wiki page that was > >>> posted earlier in this thread). > >>> > >>> > I now have several other nodes flapping with the following single > error > >>> > in > >>> > the cassandra.log > >>> > Error: Exception thrown by the agent : java.lang.NullPointerException > >>> > > >>> > I assume this is an unrelated problem? > >>> > >>> Do you have a full stack trace? > >>> > >>> -- > >>> / Peter Schuller > >> > > > > > > Just because you are no longer creating the big rows does not mean > they are no longer effecting you. For example, periodic compaction may > run on those keys. Did you delete the keys, and run a major compaction > to clear the data and tombstones? >