ES with one node


I'm in a situation with only one node on a big machine (2 CPU 12 core HT, 256 Go RAM, 40 To disk).

With this, the CPUs are only used at 800% ( 8 vCPU) and I do not have great performances (2000 EPS).

What can I do to increase this ?

Thanks & Regards

What type of disk/storage do you have? Both indexing and querying in Elasticsearch can be quite I/O intensive so disk performance is often the bottleneck. Have you looked at disk I/O and await, e.g. using iostat? If so, what does this look like?

EPS is a bad indicator of performance on its own as the amount of work that need to be performed per document depend on the size, complexity and mappings. What type of data do you have? Are you using parent-child and/or nested documents?


Thanks for your answer.

I'm using an external data storage with SCSI attachment. IO and await are very low.

Documents are logs sends by Rsyslog so with few fields, around 10. I'm not using parent-child or nested documents.

I agree with what you said about EPS but it's my client's indicator.

I have 1 shard by index, 1 index per day. I also have a backup node.

Do you have any indication of what I can test or what configuration I can try to increase my performance ?


This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.