High Load on Cassandra: Difference between revisions
Jump to navigation
Jump to search
No edit summary |
No edit summary |
||
| (One intermediate revision by one other user not shown) | |||
| Line 1: | Line 1: | ||
High | High load on Cassandra (the main domU on node00004) can be caused by tasks that run during the night (around 3 - 5am) or by an issue with the VPS. High load at 3-5am can usually be ignored, but other issues need to be examined as soon as possible. | ||
* If you can log in to Cassandra (the domU), examine the currently running processes by typing ''ps | aux'' and ''top''. | * If you can log in to Cassandra (the domU), examine the currently running processes by typing ''ps | aux'' and ''top''. | ||
* Feel free to restart any high-load processes by | * Feel free to restart any high-load processes by using ''/etc/init.d/XXX restart'' as required. These are often apache or pure-ftpd. | ||
* There will often be a load of at least 4 because of the [http://www.distributed.net/ distributed.net] process that runs across all four cores. | * There will often be a load of at least 4 because of the [http://www.distributed.net/ distributed.net] process that runs across all four cores. | ||
* If you cannot log in directly to Cassandra, log in to the dom0 as root and type ''xm console xen01''. You should then be at the Emohawk console and can run commands from there. | * If you cannot log in directly to Cassandra, log in to the dom0 as root and type ''xm console xen01''. You should then be at the Emohawk console and can run commands from there. | ||
| Line 9: | Line 9: | ||
If an issue cannot be put down to Cassandra itself, it's worth seeing what others are doing on the dom0. | If an issue cannot be put down to Cassandra itself, it's worth seeing what others are doing on the dom0. | ||
* Log in as root on the dom0 and type ''iostat -k -d -N 1 100''. This will view the I/O status each second for the next 100 seconds. | * Log in as root on the dom0 and type ''iostat -k -d -N 1 -z 100''. This will view the I/O status each second for the next 100 seconds. | ||
Latest revision as of 22:45, 6 April 2014
High load on Cassandra (the main domU on node00004) can be caused by tasks that run during the night (around 3 - 5am) or by an issue with the VPS. High load at 3-5am can usually be ignored, but other issues need to be examined as soon as possible.
- If you can log in to Cassandra (the domU), examine the currently running processes by typing ps | aux and top.
- Feel free to restart any high-load processes by using /etc/init.d/XXX restart as required. These are often apache or pure-ftpd.
- There will often be a load of at least 4 because of the distributed.net process that runs across all four cores.
- If you cannot log in directly to Cassandra, log in to the dom0 as root and type xm console xen01. You should then be at the Emohawk console and can run commands from there.
- Follow instructions at Cassandra (domU) is not responding as expected
If an issue cannot be put down to Cassandra itself, it's worth seeing what others are doing on the dom0.
- Log in as root on the dom0 and type iostat -k -d -N 1 -z 100. This will view the I/O status each second for the next 100 seconds.