mirror of
https://github.com/s3fs-fuse/s3fs-fuse.git
synced 2026-04-25 21:35:58 +03:00
[GH-ISSUE #1175] 'use_cache' flag corrupt data being read. #616
Labels
No labels
bug
bug
dataloss
duplicate
enhancement
feature request
help wanted
invalid
need info
performance
pull-request
question
question
testing
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
starred/s3fs-fuse#616
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @Teej42 on GitHub (Oct 16, 2019).
Original GitHub issue: https://github.com/s3fs-fuse/s3fs-fuse/issues/1175
Version of s3fs being used (s3fs --version)
V1.85
Version of fuse being used (pkg-config --modversion fuse, rpm -qi fuse, dpkg -s fuse)
2.9.2
Kernel information (uname -r)
4.15.0-65-generic
GNU/Linux Distribution, if applicable (cat /etc/os-release)
CentOS Linux 7 (Core)
s3fs command line used, if applicable
Working:
Broken:
Details about issue
I am using S3FS with IBM Information Server's PX Engine to communicate with IBM COS buckets. The working commands function as expected, the data are accurate, but it eats up the root directory, which is a bad thing in a Kubernetes environment (running out of
/space? K8 nuke the pod.)So I decided to try to use
use_cacheto redirect the cache to a mounted PVC. But each and every time I do so, I would get the following PX Engine error message in a random manner:To translate the error message, the first operator called
importis reading text data on partition node 22 (starts from 0), expecting a fixed length record of 590 bytes, and was only given 160 bytes. Note, this behavior is visibly random - this error messages happens at a random number of record, random number of partition node, and a random length. Perhaps with some careful review, we could nail down a pattern out of this.However, I can attest that running the working mount command would never trigger this issue. Adding
use_cachewould trigger this error each and every time I attempt to run the same job. Reverting back to the working mount command restores the working condition.I tried the
del_cacheoption, and it did not change anything. I have not tried any other options yet as I can not see anything relevant to this issue.The configuration settings are based on the review of this page: https://medium.com/@ozeri/file-like-access-to-ibm-cloud-object-storage-using-s3fs-f5094ed42594
Please advise.
@senfbrot commented on GitHub (Oct 24, 2019):
We are facing a similar problem by using Centos as well.
Version of s3fs being used (s3fs --version)
Amazon Simple Storage Service File System V1.85(commit:unknown) with OpenSSL
Version of fuse being used (pkg-config --modversion fuse, rpm -qi fuse, dpkg -s fuse)
Kernel information (uname -r)
3.10.0-1062.4.1.el7.x86_64
GNU/Linux Distribution, if applicable (cat /etc/os-release)
CentOS Linux 7 (Core)
/etc/fstab entry, if applicable
Working:
Not working:
Details about issue
When I copy a file from one bucket to another bucket through s3fs, then the resulting file is in most cases corrupt. The file size of source and target file is identical, but the checksums differ.
When
use_cacheis disabled, then there is no problem.I've tried the same on an Amazon Linux instance, there seems to be no issue with using the cache or not.
@gaul commented on GitHub (Feb 3, 2020):
Could you test with master? It has some data corruption fixes. Otherwise we need some kind of way to reproduce the error. The copy command from https://github.com/s3fs-fuse/s3fs-fuse/issues/1175#issuecomment-545897600 seems promising; does this always happen?
@Teej42 commented on GitHub (Feb 4, 2020):
What should be the build instruction for MacOS, so I can build from master to test this (and the other story I opened)? I would prefers ensuring this build stay separate from the home-brew version I have already installed:
Please let me know. Thanks!
@gaul commented on GitHub (Feb 4, 2020):
You can follow these instructions:
https://github.com/s3fs-fuse/s3fs-fuse/blob/master/COMPILATION.md
Don't run
make installat the end and you can run the binary directly fromsrc/s3fs.@Teej42 commented on GitHub (Feb 5, 2020):
Unfortunately, the instruction is not complete for the MacOS platform -
During Configure, I got the message,
No package 'libcrypto' found- and only this page would provide the solution that worked for me: https://github.com/scipr-lab/libsnark/issues/99Then when I try to
make- I getUnfortunately, there is a number of libxml/xpath.h installed all over my system:
I saw this: https://github.com/s3fs-fuse/s3fs-fuse/issues/296
But I get this:
I am not much of a MacOS developer, so any advice would be welcome.
@gaul commented on GitHub (Jun 4, 2020):
It might be easier to use the compilation tools from Homebrew. We added some additional inode sanity checks; please test again with master and reopen if symptoms persist.