ES Cloud - Running ingest pipelines on warm nodes?

Elastic Stack Elasticsearch
42

Okay, so we're making progress here... I removed the bulk_max_size, making it default to 50, but it made it worse.

So then I tried using multiple workers 4 @ 250 and that seemed to have done it! We're back in business... We must have been bottlenecking the ingest by using a single thread, which is why it worked on a single node. :man_shrugging:

Oh and for the record, I did the following after:

  • upgraded to 7.10.1
  • increased docker resources in ECS for my Filebeat container from 768 > 1024. CPU was hovering around 45%. RAM was fine with 512 utilizing only 8% of that.
  • set the refresh_rate to 10s on the three indices via index templates

Now I'm currently reducing the cluster to 2x 29G hot instances on 2 AZ's to support our HA efforts that I originally intended. I'll check on the cluster in the morning and if all is well, I'll spin up two warm nodes and test out the ILM policies (which will probably take us to Monday).

Thanks again for all your help @stephenb!

2 Likes
43

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