Graphite I/O Too Slow
We have a graphite server up and running that is currently hosting 6000 metrics at 20 second granularity. The hardware on this server is nothing to brag about. The iowait on this node is averaging above 30 seconds. Something needs to be done to reduce the rate that we are writing from the 300/sec range to the 70/sec range.
There are a couple of things that we have looked at/considered:
1. We would like to reduce our granularity to minutely data, which from testing on a vm, I think this requires us to use the whisper-resize.py . We would have to do this for 6000 metrics however, and we are not sure what the best way to go about this is. I've read that there may be a whisper-
2. I have read a little about queuing metrics for writes, I.e. Waiting for 3 data points to accumulate before you perform the I/O. Is there a way to configure this?
3. Is there another solution that we are overlooking? Is there a solid wait to optimize the I/O? Would changing the carbon.conf MAX_UPDATES_
Question information
- Language:
- English Edit question
- Status:
- Solved
- For:
- Graphite Edit question
- Assignee:
- No assignee Edit question
- Solved by:
- Brett Riotto
- Solved:
- Last query:
- Last reply: