Couchbase Server 2.0.1 is first maintenance release for Couchbase Server 2.0. This release contains number of enhancements particularly for the rebalance, indexing and compaction operations. In addition, it also includes some critical bug fixes related to system timeouts and system stability. This includes:
Improves rebalance operation time for key-value use cases.
Improves coordination of index compaction and index building that occurs during the rebalance operation. This helps significantly reduce the index file size growth during rebalance.
Changes to high and low water marks that provides more usable memory for users.
Fixes timeout problems that were seen due to Erlang VM settings.
Fixes a critical indexing problem. A document that produced a runtime failure in one view had been unavailable to all other views within the same design document.
Installation and Upgrade
Cluster Operations
The server had experienced timeouts during rebalance if views were being indexed or compacted at the same time. This resulted in the rebalance to fail. This has been fixed.
Issues: MB-6595
In the past, if you were performing an online upgrade of a 1.8 cluster to 2.0 by adding 2.0 nodes, the
dynamic configuration file on 2.0 nodes would be purged. This resulted in any XDCR remote cluster
references to be lost on the 2.0 nodes and would produce this error message "case_clause,error". This is now fixed.
Issues: MB-7568
The server had experienced random timeouts possible due to lack of asynchronous I/O threads. This caused rebalance to fail. This has been fixed.
Issues: MB-7182
Command-line Tools
There is a new setting available in cbepctl named mutation_mem_threshold. This is the amount of RAM that can be used on the server before a client begins receiving temporary out of memory errors. For more information, see 「メモリ不足エラーの為の設定変更」.
Issues: MB-7540
Indexing and Querying
In past releases of Couchbase Server 2.0, if you had a map function that failed to index, other functions in the same design document would also fail. This has been fixed.
Issues: MB-6895
During rebalance, index files were growing to an unnecessarily large size. This has been fixed.
Issues: MB-6799
There is a new REST-API call which can be used if you experience a slow rebalance while indexing and compaction are in progress. For more information, see 「コンパクション中のリバランスの調整」.
Issues: MB-7843
Cross Datacenter Replication (XDCR)
Stats for XDCR had erroneously omitted the time taken to commit items and had also displayed the incorrect number of documents/mutations checked and replicated. This is fixed.
Issues: MB-7275
When you create a replication between two clusters, you may see two error messages: "Failed to grab remote bucket info, vbucket" and "Error replicating vbucket X". Nonetheless, replication will still start and then function as expected, but the error messages may appear for some time in the Web Console. This has been fixed.
Installation and Upgrade
When you upgrade from Couchbase Server 2.0.0 to 2.0.1 on Linux the install may not replace the
file2.beam with the latest version. This will cause indexing and querying to fail.
The workaround is to install 2.0.1 and then manually restart Couchbase Server with the following commands:
sudo /etc/init.d/couchbase-server stop sudo /etc/init.d/couchbase-server start
Issues: MB-7770
If you perform an online upgrade from Couchbase Server 2.0.0 to 2.0.1, rebalance may exit with the error:
{{{linked_process_died,....,normal}, {gen_server,call, ['capi_set_view_manager-sasl', {set_vbucket_states,
Issues: MB-7771
Database Operations
In OpenVZ Linux containers, the server had crashed and restarted when you created a Couchbase bucket. This was due to an issue in the memsup process from Erlang. To fix this issue, you should upgrade to the latest version of Erlang, and have Couchbase Server use this version:
Stop Couchbase Server.
Make a copy of your original memsup then get the patch available on GitHub: memsup patch.
Place the compiled code in
/opt/membase/lib/erlang/lib/os_mon-2.2.5/priv/bin/.
Start Couchbase Server.
Issues: MB-7626
Cluster Operations
If you query a view during cluster rebalance it will fail and return the messages "error Reason: A view spec can not consist of merges exclusively" and then "no_active_vbuckets Reason: Cannot execute view query since the node has no active vbuckets." The workaround for this situation is to handle this error and retry later in your code. Alternatively the latest version of the Java SDK will automatically retry upon these errors.
Issues: MB-7661
If you want to add a new Couchbase Server 2.0.1 node to a running 2.0 cluster, and you want to refer to this node by hostname do the following on the new node:
On Linux: create a file
named ip_start in the directory
/opt/couchbase/var/lib/couchbase for the node. In this new file,
add a line provides the DNS name or IP address that you want to use to refer to the node.
On Windows: create a file
named ip_start in the directory
c:\program files\couchbase\server\var\lib\couchbase for the node. In this new file,
add a line provides the DNS name or IP address that you want to use to refer to the node.
For more information, see Couchbase Manual 2.0, Node Upgrade Process.
Issues: MB-7664
By default most Linux systems have swappiness set to 60. This will lead to overuse of disk swap with Couchbase Server. Please follow our current recommendations on swappiness and swap space, see 「スワップ領域」.
If you try to add a node during cluster rebalance, the
rebalance may fail and return the messages {detected_nodes_change
or {ns_node_disco_events. To avoid this error complete your
cluster rebalance, add the new node and then rebalance once again. This issue will
be fixed in future releases.
Issues: MB-6280
A cluster rebalance may exit and produce the error {not_all_nodes_are_ready_yet} if you perform the rebalance right after failing over a node in the cluster. You may need to wait 60 seconds after the node failover before you attempt the cluster rebalance.
This is because the failover REST API is a synchronous operation with a timeout. If it fails to complete the failover process by the timeout, the operation internally switches into a asynchronous operation. It will immediately return and re-attempt failover in the background which will cause rebalance to fail since the failover operation is still running.
Issues: MB-7168
Indexing and Querying
If you have a large amount of data in Couchbase Server and also use views, data compaction and views compaction may not automatically start. If you encounter this issue, you should change your node settings so that data and views compaction will start in paralell. In Couchbase Web Console: Select Settings | Auto-Compaction | Process Database and View compaction in paralell.
Issues: MB-7781
Cross Datacenter Replication (XDCR)
You may experience an uneven rate of replication for nodes in a cluster. This is intermittent behavior that is due to different disk drain rates at different nodes and will subside by itself over time. There is no need to change any settings on your cluster to try to resolve this behavior.
Issues: MB-7657