You are viewing a plain text version of this content. The canonical link for it is here.
Posted to notifications@couchdb.apache.org by GitBox <gi...@apache.org> on 2021/01/29 16:19:19 UTC
[GitHub] [couchdb] nickva commented on issue #3354: max document size check is inaccurate
nickva commented on issue #3354:
URL: https://github.com/apache/couchdb/issues/3354#issuecomment-769902427
+1 to have a ceiling to max_document_size
And then I believe if we see a 2103 it should be a bug and we shouldn't see the error if we do the chunking properly.
For chunking we may need a ceiling to the `binary_chunk_size` parameter which would depend on the encryption overhead. With the current default (100KB), any encryption overhead per-row is guaranteed to generate 2103 errors and it's kind of subtle to figure out why.
Is 57 bytes per-row overhead you mentioned fixed for all encryption types, or it depends on the plugin which does the actual encryption? If it depends on the plugin, maybe we should have a new aegis plugin API which exposes the per-row-encryption-overhead. Or even better, switch to encrypting, then chunking. Would that be possible? That would eliminate the involvement of chunking parameter from the discussion altogether.
> If I attempt a _bulk_docs with two documents, one of which would cause a 2103 error on the txn and one that doesn't, what should happen?
Where was the 2103 error raised for bulk_docs, would you have a stack trace? With _bulk_docs I can see getting a 2101 (transaction size exceeded) easily. There is a heuristic there how we split the request into smaller transactions currently and it probably needs adjustment.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org