incubator-cassandra-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jonathan Ellis <jbel...@gmail.com>
Subject Re: 0.7.0beta2 with 128G memory = bizarre tiny files. help?
Date Wed, 27 Oct 2010 20:54:54 GMT
Sorry, I should have been more clear: rc1 is upgrade-in-place
compatible with b2, so I would have liked to see what the settings
were before running it through an export/import cycle.

These settings look fine and I doubt you will see bizarre tiny files.

On Wed, Oct 27, 2010 at 2:52 PM, Chip Salzenberg <chip@pobox.com> wrote:
> OK, after upgrading to "apache-cassandra-2010-10-27_12-30-38", here's
> what "show keyspaces" says about the relevant keyspace:
>
>    ColumnFamily: Standard1
>      Columns sorted by: org.apache.cassandra.db.marshal.BytesType
>      Row cache size / save period: 0.0/0
>      Key cache size / save period: 200000.0/3600
>      Memtable thresholds: 9.590625/2046/60
>      GC grace seconds: 864000
>      Compaction min/max thresholds: 4/32
>
> I created it with "schematool import" off of the stock config in b2.
>
>
> On 10/27/2010 11:36 AM, Jonathan Ellis wrote:
>> Not in b2.
>> On Oct 27, 2010 11:08 AM, "Chip Salzenberg" <chip@pobox.com> wrote:
>>> Are these values available in JMX? If so, which ones are they? I'd
>>> prefer not to possibly destroy the evidence by upgrading.
>>>
>>> On 10/27/2010 6:50 AM, Jonathan Ellis wrote:
>>>> Connect with the cli and run "show keyspaces." It sounds like
>>>> Cassandra somehow picked ludicrously low memtable thresholds. There
>>>> will be a line like
>>>>
>>>> Memtable thresholds: 0.2953125/63/60
>>>>
>>>> This may not be in the b2 cli, in which case you will need to upgrade
>>>> to the nightly from
>>>>
>> http://hudson.zones.apache.org/hudson/job/Cassandra/lastSuccessfulBuild/artifact/cassandra/build/
>>>> (essentially the same as rc1 being voted on).
>>>>
>>>> On Wed, Oct 27, 2010 at 4:17 AM, Chip Salzenberg <chip@pobox.com> wrote:
>>>>> I recently tested 0.7.0beta2 with good results on a reasonably powerful
>>>>> machine: 8 Xeon cores (16 if you count hyperthreads), 64G memory, and
>>>>> some nice HP RAID. So far so good. But when I took the same config and
>>>>> moved it to a basically identical box with 128G of memory, cassandra
>>>>> started responding to writes by creating a plethora of teeny tiny
>>>>> sstables -- something it had not done before. For example:
>>>>>
>>>>> INFO [FLUSH-WRITER-POOL:1] 2010-10-27 01:53:27,881 Memtable.java (line
>>>>> 150) Writing Memtable-Standard1@950886315(651 bytes, 18 operations)
>>>>> INFO [MUTATION_STAGE:8] 2010-10-27 01:53:27,882 ColumnFamilyStore.java
>>>>> (line 459) switching in a fresh Memtable for Standard1 at
>>>>>
>> CommitLogContext(file='/var/lib/cassandra/commitlog/_chip/CommitLog-1288169534132.log',
>>>>> position=27598)
>>>>> INFO [MUTATION_STAGE:8] 2010-10-27 01:53:27,883 ColumnFamilyStore.java
>>>>> (line 771) Enqueuing flush of Memtable-Standard1@1383310803(384 bytes,
>>>>> 10 operations)
>>>>> INFO [FLUSH-WRITER-POOL:1] 2010-10-27 01:53:27,902 Memtable.java (line
>>>>> 157) Completed flushing
>>>>> /var/lib/cassandra/data/_chip/Keyspace1/Standard1-e-7-Data.db
>>>>> INFO [FLUSH-WRITER-POOL:1] 2010-10-27 01:53:27,903 Memtable.java (line
>>>>> 150) Writing Memtable-Standard1@996627145(360 bytes, 10 operations)
>>>>> INFO [MUTATION_STAGE:3] 2010-10-27 01:53:27,927 ColumnFamilyStore.java
>>>>> (line 459) switching in a fresh Memtable for Standard1 at
>>>>>
>> CommitLogContext(file='/var/lib/cassandra/commitlog/_chip/CommitLog-1288169534132.log',
>>>>> position=31446)
>>>>> INFO [MUTATION_STAGE:3] 2010-10-27 01:53:27,927 ColumnFamilyStore.java
>>>>> (line 771) Enqueuing flush of Memtable-Standard1@1909350010(957 bytes,
>>>>> 26 operations)
>>>>> INFO [FLUSH-WRITER-POOL:1] 2010-10-27 01:53:27,932 Memtable.java (line
>>>>> 157) Completed flushing
>>>>> /var/lib/cassandra/data/_chip/Keyspace1/Standard1-e-8-Data.db
>>>>> INFO [FLUSH-WRITER-POOL:1] 2010-10-27 01:53:27,933 Memtable.java (line
>>>>> 150) Writing Memtable-Standard1@1383310803(384 bytes, 10 operations)
>>>>> INFO [CompactionExecutor:1] 2010-10-27 01:53:27,934
>>>>> CompactionManager.java (line 233) Compacting
>>>>>
>> [org.apache.cassandra.io.sstable.SSTableReader(path='/var/lib/cassandra/data/_chip/Keyspace1/Standard1-e-5-Data.db'),org.apache.cassandra.io.sstable.SSTableReader(path='/var/lib/cassandra/data/_chip/Keyspace1/Standard1-e-6-Data.db'),org.apache.cassandra.io.sstable.SSTableReader(path='/var/lib/cassandra/data/_chip/Keyspace1/Standard1-e-7-Data.db'),org.apache.cassandra.io.sstable.SSTableReader(path='/var/lib/cassandra/data/_chip/Keyspace1/Standard1-e-8-Data.db')]
>>>>> INFO [MUTATION_STAGE:17] 2010-10-27 01:53:27,936 ColumnFamilyStore.java
>>>>> (line 459) switching in a fresh Memtable for Standard1 at
>>>>>
>> CommitLogContext(file='/var/lib/cassandra/commitlog/_chip/CommitLog-1288169534132.log',
>>>>> position=33074)
>>>>> INFO [MUTATION_STAGE:17] 2010-10-27 01:53:27,936 ColumnFamilyStore.java
>>>>> (line 771) Enqueuing flush of Memtable-Standard1@595066677(396 bytes,
11
>>>>> operations)
>>>>> INFO [FLUSH-WRITER-POOL:1] 2010-10-27 01:53:28,037 Memtable.java (line
>>>>> 157) Completed flushing
>>>>> /var/lib/cassandra/data/_chip/Keyspace1/Standard1-e-9-Data.db
>>>>>
>>>>> I tried lowering the available memory reported by bin/cassandra:
>>>>>
>>>>> system_memory_in_mb=`free -m | awk '/Mem:/ {print $2}'`
>>>>> [ "$system_memory_in_mb" -gt 65536 ] &&
>>>>> system_memory_in_mb=65536 #<<<< new
>>>>>
>>>>> Didn't help. I also tried maually setting the max memtable sizes:
>>>>>
>>>>> binary_memtable_throughput_in_mb: 512
>>>>> memtable_throughput_in_mb: 4096
>>>>>
>>>>> Didn't help.
>>>>>
>>>>> Help?
>>>>>
>>>>
>



-- 
Jonathan Ellis
Project Chair, Apache Cassandra
co-founder of Riptano, the source for professional Cassandra support
http://riptano.com

Mime
View raw message