Re: bugs report for 2.0.7
It looks like Cassandra will calculate free space by summing all the disks, and this will result in a big compaction file that can not fit in one disk. So any one could clarify this please? On Wed, Apr 30, 2014 at 2:09 PM, Yatong Zhang wrote: > I am using CQL 3 to create a table to store images and very image was > about 200K ~ 500K. I have 6 harddisks per node and cassandra was configured > with 6 data directories: > > data_file_directories: >> - /data1/cass >> - /data2/cass >> - /data3/cass >> - /data4/cass >> - /data5/cass >> - /data6/cass >> > > And every directory is on a standalone disk. But I just found when the > error occurred: > > [root@node5 images]# ll -hl >> total 3.6T >> drwxr-xr-x 4 root root 4.0K Jan 20 09:44 snapshots >> -rw-r--r-- 1 root root 456M Apr 30 13:42 >> mydb-images-tmp-jb-91068-CompressionInfo.db >> -rw-r--r-- 1 root root 3.5T Apr 30 13:42 mydb-images-tmp-jb-91068-Data.db >> -rw-r--r-- 1 root root0 Apr 30 13:42 >> mydb-images-tmp-jb-91068-Filter.db >> -rw-r--r-- 1 root root 2.0G Apr 30 13:42 mydb-images-tmp-jb-91068-Index.db >> > > [root@node5 images]# df -hl > Filesystem Size Used Avail Use% Mounted on > /dev/sda149G 7.5G 39G 17% / > tmpfs 7.8G 0 7.8G 0% /dev/shm > /dev/sda3 3.6T 1.3T 2.1T 38% /data1 > /dev/sdb1 3.6T 1.4T 2.1T 39% /data2 > /dev/sdc1 3.6T 466G 3.0T 14% /data3 > /dev/sdd1 3.6T 1.3T 2.2T 38% /data4 > /dev/sde1 3.6T 1.3T 2.2T 38% /data5 > /dev/sdf1 3.6T 3.6T 0 100% /data6 > > *mydb-images-tmp-jb-91068-Data.db *almost occupied all the disk space (4T > harddisk with 3.6T actual usable size) > > after I restated cassandra, very thing seems to be fine: > > -rw-r--r-- 1 root root 19K Apr 30 13:58 >> mydb_oe-images-tmp-jb-96242-CompressionInfo.db >> -rw-r--r-- 1 root root 145M Apr 30 13:58 >> mydb_oe-images-tmp-jb-96242-Data.db >> -rw-r--r-- 1 root root 64K Apr 30 13:58 >> mydb_oe-images-tmp-jb-96242-Index.db >> > > [root@node5 images]# df -hl > Filesystem Size Used Avail Use% Mounted on > /dev/sda149G 7.5G 39G 17% / > tmpfs 7.8G 0 7.8G 0% /dev/shm > /dev/sda3 3.6T 1.3T 2.1T 38% /data1 > /dev/sdb1 3.6T 1.4T 2.1T 39% /data2 > /dev/sdc1 3.6T 466G 3.0T 14% /data3 > /dev/sdd1 3.6T 1.3T 2.2T 38% /data4 > /dev/sde1 3.6T 1.3T 2.2T 38% /data5 > /dev/sdf1 3.6T 662M 3.4T 1% /data6 > > So my questions are: > > 1. I am using CQL3 and is there a limit for 'tables' created by CQL3? > 2. I specified 6 data directories with each on a stand alone disk, is it > OK? > 3. Why the tmp db file is so large? Is it normal or a bug? > > > So could any one please help to solve this issue? Any help is of great > appreciation and thanks a lot! > > > On Wed, Apr 30, 2014 at 12:04 PM, Yatong Zhang wrote: > >> Thanks for the response. I've checked the system logs and harddisk smartd >> info, and no errors found. Any hints to locate the problem? >> >> >> On Wed, Apr 30, 2014 at 9:26 AM, Michael Shuler >> wrote: >> >>> Then you likely need to fix your I/O problem. The most recent error you >>> posted is an EOFException - the file being read ended unexpectedly. >>> Probably when you ran out of disk space. >>> >>> -- >>> Michael >>> >>> >>> On 04/29/2014 07:48 PM, Yatong Zhang wrote: >>> Here is another type of exception, seems all are I/O related: INFO [SSTableBatchOpen:1] 2014-04-29 14:44:35,548 SSTableReader.java (line > 223) Opening > /data2/cass/system/compaction_history/system-compaction_ > history-jb-6956 > (447252 bytes) > INFO [SSTableBatchOpen:2] 2014-04-29 14:44:35,553 SSTableReader.java > (line 223) Opening > /data2/cass/system/compaction_history/system-compaction_ > history-jb-6958 > (257 bytes) > INFO [SSTableBatchOpen:3] 2014-04-29 14:44:35,554 SSTableReader.java > (line 223) Opening > /data2/cass/system/compaction_history/system-compaction_ > history-jb-6957 > (257 bytes) > INFO [main] 2014-04-29 14:44:35,592 ColumnFamilyStore.java (line 248) > Initializing system.batchlog > INFO [main] 2014-04-29 14:44:35,596 ColumnFamilyStore.java (line 248) > Initializing system.sstable_activity > INFO [SSTableBatchOpen:1] 2014-04-29 14:44:35,601 SSTableReader.java > (line 223) Opening > /data2/cass/system/sstable_activity/system-sstable_activity-jb-8084 > (1562 > bytes) > INFO [SSTableBatchOpen:2] 2014-04-29 14:44:35,604 SSTableReader.java > (line 223) Opening > /data2/cass/system/sstable_activity/system-sstable_activity-jb-8083 > (2075 > bytes) > INFO [SSTableBatchOpen:3] 2014-04-29 14:44:35,605 SSTableReader.java > (line 223) Opening > /data2/cass/system/sstable_activity/system-sstable_activity-jb-8085 > (1555 > bytes) > INFO [main] 2014-04-29 14:44:35,687 AutoSavingCache.java (line 114) > reading saved cache
Next-generation Cassandra Conference, June 13, Austin TX
I'm putting together a conference for Cassandra committers, driver authors, and serious not-yet- committers. Space is limited to 30, so I'm really sincere about this not competing with the Cassandra Summit [1] as a user-level event. The format will be a single track of 40 minute talks, with 20 minutes of open discussion after each. Topics could include things like - Implementing alternative compaction strategies - Migrating memtables off-heap - Adding user-defined types to Cassandra drivers - RAMP transactions - Server support for top-K queries - etc. DataStax will provide breakfast, lunch, and dinner from local restaurants [2]. The schedule will also include two thirty minute breaks during the day to provide opportunities for unstructured discussion. We expect that one of the main benefits of attendance will be the opportunity to visit and network with your peers this way. Venue and combined CFP/registration info are at http://learn.datastax.com/Next-GenerationCassandraConference.html. It's a bit spartan but it should get the job done. [1] http://www.cvent.com/events/cassandra-summit-2014/event-summary-176f061a4b144525ae05f18cd9a9cb12.aspx [2] I think we're actually contractually obligated to buy some food from the hotel. I'll minimize this to the extent possible. -- Jonathan Ellis Project Chair, Apache Cassandra co-founder, http://www.datastax.com @spyced
Re: Next-generation Cassandra Conference, June 13, Austin TX
Hi Jonathan, Are you planning to stream the event or to record the talks and make them available later? I am really interested in the Cassandra internals and eventually submitting some code, but I am not yet at the level where I could contribute to an event like this. I would be really interested in most of the topics you listed above. Best regards, Clint On Wed, Apr 30, 2014 at 2:38 PM, Jonathan Ellis wrote: > I'm putting together a conference for Cassandra committers, driver > authors, and serious not-yet- committers. Space is limited to 30, so > I'm really sincere about this not competing with the Cassandra Summit > [1] as a user-level event. > > The format will be a single track of 40 minute talks, with 20 minutes > of open discussion after each. Topics could include things like > > - Implementing alternative compaction strategies > - Migrating memtables off-heap > - Adding user-defined types to Cassandra drivers > - RAMP transactions > - Server support for top-K queries > - etc. > > DataStax will provide breakfast, lunch, and dinner from local > restaurants [2]. The schedule will also include two thirty minute breaks > during the day to provide opportunities for unstructured discussion. > We expect that one of the main benefits of attendance will be the > opportunity to visit and network with your peers this way. > > Venue and combined CFP/registration info are at > http://learn.datastax.com/Next-GenerationCassandraConference.html. > It's a bit spartan but it should get the job done. > > [1] > http://www.cvent.com/events/cassandra-summit-2014/event-summary-176f061a4b144525ae05f18cd9a9cb12.aspx > > [2] I think we're actually contractually obligated to buy some food > from the hotel. I'll minimize this to the extent possible. > > -- > Jonathan Ellis > Project Chair, Apache Cassandra > co-founder, http://www.datastax.com > @spyced >
Re: Next-generation Cassandra Conference, June 13, Austin TX
I wish we could, but it's more than I can bite off logistically this year. If it's as productive as I think it will be, we'll up our game and add recording for next year. On Wed, Apr 30, 2014 at 4:44 PM, Clint Kelly wrote: > Hi Jonathan, > > Are you planning to stream the event or to record the talks and make them > available later? I am really interested in the Cassandra internals and > eventually submitting some code, but I am not yet at the level where I > could contribute to an event like this. I would be really interested in > most of the topics you listed above. > > Best regards, > Clint > > > > > On Wed, Apr 30, 2014 at 2:38 PM, Jonathan Ellis wrote: > >> I'm putting together a conference for Cassandra committers, driver >> authors, and serious not-yet- committers. Space is limited to 30, so >> I'm really sincere about this not competing with the Cassandra Summit >> [1] as a user-level event. >> >> The format will be a single track of 40 minute talks, with 20 minutes >> of open discussion after each. Topics could include things like >> >> - Implementing alternative compaction strategies >> - Migrating memtables off-heap >> - Adding user-defined types to Cassandra drivers >> - RAMP transactions >> - Server support for top-K queries >> - etc. >> >> DataStax will provide breakfast, lunch, and dinner from local >> restaurants [2]. The schedule will also include two thirty minute breaks >> during the day to provide opportunities for unstructured discussion. >> We expect that one of the main benefits of attendance will be the >> opportunity to visit and network with your peers this way. >> >> Venue and combined CFP/registration info are at >> http://learn.datastax.com/Next-GenerationCassandraConference.html. >> It's a bit spartan but it should get the job done. >> >> [1] >> http://www.cvent.com/events/cassandra-summit-2014/event-summary-176f061a4b144525ae05f18cd9a9cb12.aspx >> >> [2] I think we're actually contractually obligated to buy some food >> from the hotel. I'll minimize this to the extent possible. >> >> -- >> Jonathan Ellis >> Project Chair, Apache Cassandra >> co-founder, http://www.datastax.com >> @spyced >> -- Jonathan Ellis Project Chair, Apache Cassandra co-founder, http://www.datastax.com @spyced
