Open xlzheng021 opened 3 years ago
So, I'd like to first ask more information about the environment. Specifically, how much ram is available to the process? And is this a 64-bit system?
It would be highly unusual for an index with so few documents to run out of memory, as we regularly test with index built on 5 million documents. That said, memory usage does vary by the type of index you are building. Do you have many fields? Or have large values in those fields? The reason I'm asking is that sometimes people end up indexing large values accidentally (usually some binary data inside a string that gets index as a single, unusually large token)
You asked about Bleve v2, and certainly I do recommend upgrading, but we have not fixed any issues that are obviously related to this behavior.
As for next steps, there is no specific setting to control memory used by the indexing. At Couchbase they will throttle the rate of indexing if the memory usage is high, but this done by the application, not bleve itself. And if the memory is consumed during a merge as you suggest, it likely would not help this case. So, that leaves us with either you being able to share some program and data that reproduce the problem, or digging in yourself to see where and why the memory is used.
Hi,
I am using the Bleve 1.0.14 with scorch indexer, it seems the indexer is aggressively using the whole memory that I had during the final indexing merge, which caused the "out of memory" issue, is there any way that we could limit the memory usage during the index time? Or could upgrade to Bleve 2.* fix the issue?
Thanks,
The indexer config that we are using is:
The documentation nubmer we are dealing with is about 55k docs.