Corosync/CTDB and LVS performance issues

This document (7006849) is provided subject to the disclaimer at the end of this document.

Environment

SUSE Linux Enterprise Server 11 Service Pack 1

Situation

Using a Corosync/LVS and CTDB setup may show some performance issues.

Many small files are written via smbclient in short intervals to a share.

In /var/log/messages the following error could be reported:

NT_STATUS_IO_TIMEOUT opening remote file \testfile302
Receiving SMB: Server stopped responding
Receiving SMB: Server stopped responding
NT_STATUS_IO_TIMEOUT opening remote file \testfile304
...

Reason: Using the provided CTDB agent the CTDB_LVS_PUBLIC_IP option is not written to the /etc/sysconfig/ctdb.

Resolution

Since the /etc/sysconfig/ctdb will be overwritten each time the agent starts, editing this file makes no sense.
Instead backup the agent script ( /usr/lib/ocf/resource.d/heartbeat/CTDB ) and add the following option after
the following line:

CTDB_SERVICE_WINBIND=winbind

CTDB_LVS_PUBLIC_IP=xxx.xxx.xxx.xxx/yy

where xxx.xxx.xxx.xxx is the public ip adress of the LVS and yy ist the netmask in CIDR notation.

Stop the CTDB resource and start it again. Check the contents of /etc/sysconfig/ctdb

Additional Information

Reproduction: Write many small files via smbclient in short intervals to a share.

You may get messages as following: ...

NT_STATUS_IO_TIMEOUT opening remote file \testfile302
Receiving SMB: Server stopped responding
Receiving SMB: Server stopped responding
NT_STATUS_IO_TIMEOUT opening remote file \testfile304
...

Reason: Using the provided CTDB agent the CTDB_LVS_PUBLIC_IP option is not written to the /etc/sysconfig/ctdb.

Disclaimer

This Support Knowledgebase provides a valuable tool for SUSE customers and parties interested in our products and solutions to acquire information, ideas and learn from one another. Materials are provided for informational, personal or non-commercial use within your organization and are presented "AS IS" WITHOUT WARRANTY OF ANY KIND.