System call filter/OnError and OnOutOfMemoryError - Bootstrap check

As part of ES 5.x bootstrap, it checks for System call filter. What actually is System call filter

Also can you please explain whats OnError and OnOutOfMemoryError check

May be read https://www.elastic.co/guide/en/elasticsearch/reference/current/bootstrap-checks.html

Thanks @dadoonet . Most of these checks are for Linux machines. I dont see much of checks for windows servers( if we want to run ES on windows servers) . Wht's the recommend OS for running ES. Is it Linux servers or windows servers. Would there be much of performance difference running on a windows server vs linux server assuming both have the same hardware specs

Windows is OK. But I don't have much experience on running Elasticsearch on Windows.

Would there be much of performance difference running on a windows server vs linux server assuming both have the same hardware specs?

Good question. I don't think we have benchmarks like this. @danielmitterdorfer are you aware of this?
I believe that as soon your system does not swap to disk in both cases and that elasticsearch is the only "main" process running, it should be similar... But that's just a guess.

Hi,

At the moment, the Elasticsearch development team runs its regular benchmarks exclusively on Linux. There may be some (irregular) benchmarks on Windows. However, I am not aware of any OTOH. Anyway, benchmarks should always be created based on your data set and traffic patterns in order to allow you to draw any useful conclusions.

For our nightly benchmarks, we use Rally (I'm the main author of this tool). You can use it to benchmark a (remote) Windows server but the tool itself requires a Unix machine. So you could place your load generator (Rally) on a Unix machine and run your benchmarks against Windows and Linux machines. See our docs for an example how to benchmark an existing cluster with Rally.

Daniel

Thanks for the info

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