#s3
49 posts · Last used Sep 24
Playing with #NetBSD on a hand-me-down Dell Inspiron with a spinny disc.
So far, it's a joy to use, even though #S3 #Suspend isn't working.
The spinny disc sounds are actually kinda therapeutic XD
`-/oshdmNMNdhyo+:-`
y/s+:- `.-:+oydNMMMMNhs/- OS: NetBSD 11.0_RC7 amd64
-m+NMMMMMMMMMMMMMMMMMMMNdhmNMMMmdhs+/- Uptime: 8 hours, 12 mins -m+NMMMMMMMMMMMMMMMMMMMMmy+: Packages: 124 (pkg_info)
-N/dMMMMMMMMMMMMMMMds: Shell: ksh v5.2.14 99/07/13.2 -N/hMMMMMMMMMmho: Terminal: /dev/pts/5
-N/-:/++/:.` CPU: Intel 686-class (4)
:M+ Memory: 16234MiB / 16281MiB
:Mo
:Ms
:Ms
:Ms
:Ms
:Ms
:Ms
:Ms
:Ms
Dear #MastoAdmin for anyone using #S3 storage I would highly recommend disabling S3 multipart uploads for a day (or forever) to see if you notice any improvements.
Pre-existing issues with Mastodon and S3, plus some recent regressions in the #AWS S3 library can cause a variety of issues, notably in the ingress #sidekiq worker, such as:
- infinite disk usage in /tmp
- higher outbound bandwidth use (2-4x)
- higher cpu and memory use by sidekiq
- higher costs on providers like Wasabi that have minimum duration charges for incomplete multipart uploads
You can do this by setting S3_MULTIPART_THRESHOLD=1073741824 (any sufficiently high value that all processed media will fit below).
Note that setting S3_FORCE_SINGLE_REQUEST=true is not sufficient, as this only impacts reading files from S3 to disk, and not uploading to S3. Increasing S3 timeouts and retries may help, but doesn't solve the issues entirely.
The results will depend on your environment and specific S3 provider, but are significant and worth testing to see if they affect you.
https://github.com/mastodon/mastodon/issues/39863
"There is a new release of SeaweedFS. Happy updating your instance and don't forget to check the changelog."
4.22Comments For clusters with multiple volume servers with multiple disks, and if Erasure Coding is used, please upgrade! #9219 fixed a problem when the volume server is restarted, some EC shards may be reported as missing while actually they are...
![[GLOBAL][Storage] - Object Storage S3 maintenance notification](https://dka575ofm4ao0.cloudfront.net/pages-twitter_logos/original/164468/OVHcloud_Logo_630x420.png)