Open xrefft opened 4 years ago
We believe we're facing the same issue once we have started using v1.89. Did you figure out a solution/root-cause/workaround for this @xrefft ?
Mount commands we have tried:
1. s3fs -o ecs -o retries=20 -o max_dirty_data=100 -o ensure_diskfree=200 <BUCKET> <MOUNT_POINT>
2. s3fs -o ecs -o retries=20 -o max_dirty_data=50 -o ensure_diskfree=200 <BUCKET> <MOUNT_POINT>
You should note that while we have not specified any disk cache in either command above, we create the /tmp
as a tmpfs before running the above commands like this (so as to avoid disk activity on ec2 ebs volumes):
mount -t tmpfs -o size=400m tmpfs /tmp
The latest version of s3fs reports IO Error when continuous appending to one single file.
we noticed
[ERR] curl.cpp:RequestPerform(2596): ### CURLE_GOT_NOTHING
error in curldbg log. which probably means concurrency problem.Additional Information
Version of s3fs being used (s3fs --version)
It happend on s3fs V1.87 (commit:8b7dd82) with OpenSSL. Commit #1448 .
Version of fuse being used (pkg-config --modversion fuse, rpm -qi fuse, dpkg -s fuse)
Kernel information (uname -r)
3.10.0-957.27.2.el7.x86_64
GNU/Linux Distribution (cat /etc/os-release)
s3fs command line used
s3fs syslog messages (grep s3fs /var/log/syslog, journalctl | grep s3fs, or s3fs outputs)
Details about issue
The latest version of s3fs reports IO Error when continuous appending to one single file.
we noticed
[ERR] curl.cpp:RequestPerform(2596): ### CURLE_GOT_NOTHING
error in curldbg log. which probably means concurrency problem. PS:When using AWS SDK, OSS will not report this error, but using s3fs will cause this error.Python test code we use is: