Re: Cassandra shuts down; was:Cassandra crashes
Ideally, you should get back pressure in the form of dropped messages before you see crashes, but if turning down the heap allocation was the only thing you did, there are other changes required (several mentioned by Romain above are very good places to start). A few other ideas: - did you adjust ParNew along with heap? - you may want to adjust SurvivorRatio and MaxTenuringThreshold (change both to 4 as a starting point) in JVM - definitely play with compaction throughput by turning it way up since you have IO capacity These will cause you to GC and compact continuously in this environment, but it should at least keep going On Wed, Sep 4, 2013 at 9:14 AM, Romain HARDOUIN wrote: > Have you tried to tweak settings like memtable_total_space_in_mb and > flush_largest_memtables_at? > Also, the compaction manager seems to be pretty busy, take a look at > in_memory_compaction_limit_in_mb. > And with SSD hardware you should modifiy multithreaded_compaction, > compaction_throughput_mb_per_sec, concurrent_reads and concurrent_writes. > Of course 2GB of RAM is low but tweak these settings might help you. > Maybe some guru could confirm/infirm that. > > > > > De :Jan Algermissen > A :user@cassandra.apache.org, > Date : 04/09/2013 12:29 > Objet : Re: Cassandra shuts down; was:Cassandra crashes > -- > > > > Romain, > > > On 04.09.2013, at 11:11, Romain HARDOUIN > wrote: > > > Maybe you should include the end of Cassandra logs. > > There is nothing that seems interesting in cassandra.log. Below you find > system.log. > > > What comes to my mind when I read your first post is OOM killer. > > But what you describe later is not the case. > > Just to be sure, have you checked /var/log/messages? > > Nothing there, just occasional Firewall TCP rejections. > > Somehow I think I am simply overloading the whole cluster (see the hinted > handoff messages in the log). Could that be due to the limited memory of > 2GB my nodes have? IOW, not enough space to buffer up the writes before > dumping to disk? > > Also, my overall write performance is actually pretty bad compared to what > I read about C*. Before I thought it was the client doing to much work or > the network. Turns out that's not the case. > > I'd expect C* to sort of just suck in my rather small amount of data - > must be me, not using the right configuration. Oh well, I'll get there :-) > Thanks anyhow. > > Jan > > > > >
Re: Cassandra shuts down; was:Cassandra crashes
Have you tried to tweak settings like memtable_total_space_in_mb and flush_largest_memtables_at? Also, the compaction manager seems to be pretty busy, take a look at in_memory_compaction_limit_in_mb. And with SSD hardware you should modifiy multithreaded_compaction, compaction_throughput_mb_per_sec, concurrent_reads and concurrent_writes. Of course 2GB of RAM is low but tweak these settings might help you. Maybe some guru could confirm/infirm that. De :Jan Algermissen A : user@cassandra.apache.org, Date : 04/09/2013 12:29 Objet : Re: Cassandra shuts down; was:Cassandra crashes Romain, On 04.09.2013, at 11:11, Romain HARDOUIN wrote: > Maybe you should include the end of Cassandra logs. There is nothing that seems interesting in cassandra.log. Below you find system.log. > What comes to my mind when I read your first post is OOM killer. > But what you describe later is not the case. > Just to be sure, have you checked /var/log/messages? Nothing there, just occasional Firewall TCP rejections. Somehow I think I am simply overloading the whole cluster (see the hinted handoff messages in the log). Could that be due to the limited memory of 2GB my nodes have? IOW, not enough space to buffer up the writes before dumping to disk? Also, my overall write performance is actually pretty bad compared to what I read about C*. Before I thought it was the client doing to much work or the network. Turns out that's not the case. I'd expect C* to sort of just suck in my rather small amount of data - must be me, not using the right configuration. Oh well, I'll get there :-) Thanks anyhow. Jan
Re: Cassandra shuts down; was:Cassandra crashes
Romain, On 04.09.2013, at 11:11, Romain HARDOUIN wrote: > Maybe you should include the end of Cassandra logs. There is nothing that seems interesting in cassandra.log. Below you find system.log. > What comes to my mind when I read your first post is OOM killer. > But what you describe later is not the case. > Just to be sure, have you checked /var/log/messages? Nothing there, just occasional Firewall TCP rejections. Somehow I think I am simply overloading the whole cluster (see the hinted handoff messages in the log). Could that be due to the limited memory of 2GB my nodes have? IOW, not enough space to buffer up the writes before dumping to disk? Also, my overall write performance is actually pretty bad compared to what I read about C*. Before I thought it was the client doing to much work or the network. Turns out that's not the case. I'd expect C* to sort of just suck in my rather small amount of data - must be me, not using the right configuration. Oh well, I'll get there :-) Thanks anyhow. Jan > > Romain > INFO [ScheduledTasks:1] 2013-09-04 07:17:09,057 StatusLogger.java (line 96) KeyCache216 936 all INFO [ScheduledTasks:1] 2013-09-04 07:17:09,057 StatusLogger.java (line 102) RowCache 00 all org.apache.cassandra.cache.SerializingCacheProvider INFO [ScheduledTasks:1] 2013-09-04 07:17:09,058 StatusLogger.java (line 109) ColumnFamilyMemtable ops,data INFO [ScheduledTasks:1] 2013-09-04 07:17:09,082 StatusLogger.java (line 112) system.local 4,52 INFO [ScheduledTasks:1] 2013-09-04 07:17:09,083 StatusLogger.java (line 112) system.peers 0,0 INFO [ScheduledTasks:1] 2013-09-04 07:17:09,083 StatusLogger.java (line 112) system.batchlog 0,0 INFO [ScheduledTasks:1] 2013-09-04 07:17:09,083 StatusLogger.java (line 112) system.NodeIdInfo 0,0 INFO [ScheduledTasks:1] 2013-09-04 07:17:09,083 StatusLogger.java (line 112) system.LocationInfo 0,0 INFO [ScheduledTasks:1] 2013-09-04 07:17:09,084 StatusLogger.java (line 112) system.Schema 0,0 INFO [ScheduledTasks:1] 2013-09-04 07:17:09,084 StatusLogger.java (line 112) system.Migrations 0,0 INFO [ScheduledTasks:1] 2013-09-04 07:17:09,084 StatusLogger.java (line 112) system.schema_keyspaces 0,0 INFO [ScheduledTasks:1] 2013-09-04 07:17:09,084 StatusLogger.java (line 112) system.schema_columns 0,0 ERROR [FlushWriter:6] 2013-09-04 07:17:09,210 CassandraDaemon.java (line 192) Exception in thread Thread[FlushWriter:6,5,main] java.lang.OutOfMemoryError: Java heap space at org.apache.cassandra.io.util.FastByteArrayOutputStream.expand(FastByteArrayOutputStream.java:104) at org.apache.cassandra.io.util.FastByteArrayOutputStream.write(FastByteArrayOutputStream.java:220) at java.io.DataOutputStream.write(DataOutputStream.java:107) at org.apache.cassandra.io.util.DataOutputBuffer.write(DataOutputBuffer.java:60) at org.apache.cassandra.utils.ByteBufferUtil.write(ByteBufferUtil.java:328) at org.apache.cassandra.utils.ByteBufferUtil.writeWithLength(ByteBufferUtil.java:315) at org.apache.cassandra.db.ColumnSerializer.serialize(ColumnSerializer.java:55) at org.apache.cassandra.db.ColumnSerializer.serialize(ColumnSerializer.java:30) at org.apache.cassandra.db.OnDiskAtom$Serializer.serializeForSSTable(OnDiskAtom.java:62) at org.apache.cassandra.db.ColumnIndex$Builder.add(ColumnIndex.java:181) at org.apache.cassandra.db.ColumnIndex$Builder.build(ColumnIndex.java:133) at org.apache.cassandra.io.sstable.SSTableWriter.append(SSTableWriter.java:185) at org.apache.cassandra.db.Memtable$FlushRunnable.writeSortedContents(Memtable.java:489) at org.apache.cassandra.db.Memtable$FlushRunnable.runWith(Memtable.java:448) at org.apache.cassandra.io.util.DiskAwareRunnable.runMayThrow(DiskAwareRunnable.java:48) at org.apache.cassandra.utils.WrappedRunnable.run(WrappedRunnable.java:28) at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615) at java.lang.Thread.run(Thread.java:724) INFO [ScheduledTasks:1] 2013-09-04 07:17:09,210 StatusLogger.java (line 112) system.schema_columnfamilies 0,0 INFO [ScheduledTasks:1] 2013-09-04 07:17:09,218 StatusLogger.java (line 112) system.IndexInfo 0,0 INFO [ScheduledTasks:1] 2013-09-04 07:17:09,218 StatusLogger.java (line 112) system.range_xfers
Re: Cassandra shuts down; was:Cassandra crashes
Maybe you should include the end of Cassandra logs. What comes to my mind when I read your first post is OOM killer. But what you describe later is not the case. Just to be sure, have you checked /var/log/messages? Romain De :Jan Algermissen A : user@cassandra.apache.org, Date : 04/09/2013 10:52 Objet : Re: Cassandra shuts down; was:Cassandra crashes The subject line isn't appropriate - the servers do not crash but shut down. Since the log messages appear several lines before the end of the log file, I only saw afterwards. Excuse the confusion. Jan On 04.09.2013, at 10:44, Jan Algermissen wrote: > Hi, > > I have set up C* in a very limited environment: 3 VMs at digitalocean with 2GB RAM and 40GB SSDs, so my expectations about overall performance are low. > > Keyspace uses replication level of 2. > > I am loading 1.5 Mio rows (each 60 columns of a mix of numbers and small texts, 300.000 wide rows effektively) in a quite 'agressive' way, using java-driver and async update statements. > > After a while of importing data, I start seeing timeouts reported by the driver: > > com.datastax.driver.core.exceptions.WriteTimeoutException: Cassandra timeout during write query at consistency ONE (1 replica were required but only 0 acknowledged the write > > and then later, host-unavailability exceptions: > > com.datastax.driver.core.exceptions.UnavailableException: Not enough replica available for query at consistency ONE (1 required but only 0 alive). > > Looking at the 3 hosts, I see two C*s went down - which explains that I still see some writes succeeding (that must be the one host left, satisfying the consitency level ONE). > > > The logs tell me AFAIU that the servers shutdown due to reaching the heap size limit. > > I am irritated by the fact that the instances (it seems) shut themselves down instead of limiting their amount of work. I understand that I need to tweak the configuration and likely get more RAM, but still, I would actually be satisfied with reduced service (and likely more timeouts in the client). Right now it looks as if I would have to slow down the client 'artificially' to prevent the loss of hosts - does that make sense? > > Can anyone explain whether this is intended behavior, meaning I'll just have to accept the self-shutdown of the hosts? Or alternatively, what data I should collect to investigate the cause further? > > Jan > > > > >
Re: Cassandra shuts down; was:Cassandra crashes
The subject line isn't appropriate - the servers do not crash but shut down. Since the log messages appear several lines before the end of the log file, I only saw afterwards. Excuse the confusion. Jan On 04.09.2013, at 10:44, Jan Algermissen wrote: > Hi, > > I have set up C* in a very limited environment: 3 VMs at digitalocean with > 2GB RAM and 40GB SSDs, so my expectations about overall performance are low. > > Keyspace uses replication level of 2. > > I am loading 1.5 Mio rows (each 60 columns of a mix of numbers and small > texts, 300.000 wide rows effektively) in a quite 'agressive' way, using > java-driver and async update statements. > > After a while of importing data, I start seeing timeouts reported by the > driver: > > com.datastax.driver.core.exceptions.WriteTimeoutException: Cassandra timeout > during write query at consistency ONE (1 replica were required but only 0 > acknowledged the write > > and then later, host-unavailability exceptions: > > com.datastax.driver.core.exceptions.UnavailableException: Not enough replica > available for query at consistency ONE (1 required but only 0 alive). > > Looking at the 3 hosts, I see two C*s went down - which explains that I still > see some writes succeeding (that must be the one host left, satisfying the > consitency level ONE). > > > The logs tell me AFAIU that the servers shutdown due to reaching the heap > size limit. > > I am irritated by the fact that the instances (it seems) shut themselves down > instead of limiting their amount of work. I understand that I need to tweak > the configuration and likely get more RAM, but still, I would actually be > satisfied with reduced service (and likely more timeouts in the client). > Right now it looks as if I would have to slow down the client 'artificially' > to prevent the loss of hosts - does that make sense? > > Can anyone explain whether this is intended behavior, meaning I'll just have > to accept the self-shutdown of the hosts? Or alternatively, what data I > should collect to investigate the cause further? > > Jan > > > > >