Knowledgebase

jbd2 journal high i/o

Posted by devonblzx, 04-23-2012, 02:18 PM

I know the jbd2 is the journaling process for ext4, however, I do not ever remember having such an issue with i/o from journaling. Ever since upgrading to CentOS 6, we have had higher CPU wait than usual. When running iotop on multiple servers, most of the time, I see jbd2 using the majority of the i/o. Sometimes as much as 90% I/O. These servers are running the OpenVZ kernel, not sure if that would have an effect or not. For example: Anyone ever dealt with or seen this before? Any techniques for reducing the effect of the journal? I mean it seems like the journal is doing more work than the rest of the system. I tried raising the commit time on the file system to 60 from 5, but the I/O still exists.

Posted by ehenke, 04-30-2012, 07:52 AM

Try identify where the I/O originates, maybe it is not jbd2 inherent. Compare iotop results with commands like dstat --aio --io --disk --tcp --top-io-adv --top-bio-adv or atop -dl

Add to Favourites

Print this Article