Listing all files under specific prefix is very slow about mountpoint-s3 HOT 6 CLOSED

Hi, thanks for opening this issue. This is expected, as for Mountpoint when you do a list we have to paginate and do multiple listObject requests, getting 1000 objects in each request which can lead to this latency. Using the date as a prefix should help reduce time for listing though, as it will mean that Mountpoint can complete the listing with fewer requests.

from mountpoint-s3.

@ahmarsuhail I have a "folder" with over 96882 "files" and listing takes. I compared different s3 mount implementations s3fs-fuse - about 5 minutes goofys - 35 seconds mountpoint-s3 45 minutes!!!

There's obviously something wrong with the listing implementation in mountpoint-s3

Hey @VladimirPchelko, please take a look at the troubleshooting section for this scenario: https://github.com/awslabs/mountpoint-s3/blob/main/doc/TROUBLESHOOTING.md#slow-metadata-operations

Likely, metadata caching can help you: https://github.com/awslabs/mountpoint-s3/blob/main/doc/CONFIGURATION.md#metadata-cache

If you're still experiencing issues, please do open a new issue and we can take a look. https://github.com/awslabs/mountpoint-s3/issues/new?assignees=&labels=bug&projects=&template=bug-report.yml

from mountpoint-s3.

@ahmarsuhail
I have a "folder" with over 96882 "files" and listing takes.
I compared different s3 mount implementations
s3fs-fuse - about 5 minutes
goofys - 35 seconds
mountpoint-s3 45 minutes!!!

There's obviously something wrong with the listing implementation in mountpoint-s3

from mountpoint-s3.

@dannycjones
I need that when deploying a container there are no delays on the first request! Do you offer caching? What's the point?

I'll check, but I seriously doubt it... wait 45 minutes again.

from mountpoint-s3.

after enabling caching

real 1m27.568s
but still slower than goofys

ok

from mountpoint-s3.

@VladimirPchelko we'll be happy to look into the performance issue you're facing, but we will need more information.

Details about the size of each object, and the structure of the S3 bucket would be helpful. For example, how deeply nested are the objects within prefixes - a/file.txt or a/b/c/d/e/file.txt.

Please do open a new bug report to describe the issue and share some logs, and we can investigate.

from mountpoint-s3.