4,395 karma · joined November 23, 2015
Hardware errors are also possible but there has been a few software bugs so worth checking.
Here's a really curly one we recently found and solved at work (only in March) that causes both scrub errors and sometimes bluefs aborts, it's a kernel patch. Likely to happen under memory pressure:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/lin...
https://tracker.ceph.com/issues/22464#note-50
But there have been various others.
For erasure coded setups which is not really suited to block storage but mainly object storage via radosgw(s3) or cephfs you need minimum k+m and realistically k+m+1. That would translate to 6 minimum but realistically 7 nodes for k=4,m=2. That’s 4 data chunks and 2 redundant chunks which means you use 1.5x the storage of the raw data (half that of a replicated setup). You can do k=2,m=1 also. So 4 nodes into that case.
By default it “scrubs” basic metadata daily and does a deep scrub where it fully reads the object and confirms the checksum is correct from all 3 replicas weekly for all of the data in the cluster.
It’s automatic and enabled by default.
If each pis Ethernet was broken out individually and you used a 10G uplink switch or multiple 1G client ports then you could do better.
The write speed being lower than the read speed will be because writes have to be replicated to two other nodes in the ceph cluster (everything has 3 replicas) which are also sharing bandwidth on those same 1G links. Reads don’t need to replicate so can consume the full bandwidth.
So basically it’s all network limited for this use case. Needs a 2.5G uplink, LACP link aggregation or individual Ethernet ports to do better.
Some of these addresses were the unique style you mentioned.
So I guess you’re just lucky.
“Hey, I told you not to call me on this wall. This is an unlisted wall!”
Reusable. Seals properly. Incredible breathability. Doesn’t hurt your ears.
Main downside is speech is not the most amazing. You also get a bit of condensation.
Love mine. Also cuts the waste and has no elastic to trap the birds
https://news.ycombinator.com/item?id=31664688 https://twitter.com/marcan42/status/1534053625110351872 https://twitter.com/never_released/status/153412764108259328...
Michel Van Bizen has hundreds (thousands?6 of university level lectures on many topics science and math: https://youtube.com/c/MichelvanBiez
Discovered that channel trying to understand all modes of heat transfer through windows. Covered under thermodynamics. https://youtu.be/r3tC-ocdVoQ
Most compressed filesystems will fail to dedup stuff because for various performance related reasons the compression window is usually quite small maybe up to 128kb and the files are often not sorted in the archive such that small files would manage to compress anyway in that window.
EDIT: Apparently squashfs does include file de-dupe however DwarFS is fixing what I basically said above. It sorts files in order of similarity so that you can compress accross file boundaries as well as just removing complete duplicates. That's pretty cool.
"Clustering of files by similarity using a similarity hash function. This makes it easier to exploit the redundancy across file boundaries."
This kindof thing is more difficult to do on a writable filesystem though not impossible.
I remember that apparently for desktop applications icons to work it had to launch a process just to be each icon so was quite memory intensive. I seem to recall it also requiring “win32s”
My machine would throw a text mode confirmation prompt over the win95 graphical installer when it tried to write the boot sector as boot sector virus protection and this would freeze the system so windows 95 installation never worked.
It sure was great only using sound from one application at once.
Having said that, they now enabled Picture-in-Picture on iOS (though last check it's still in beta I joined the beta a while back) and one of Apple's requirements is you cannot but OS features (such as Picture in Picture) behind an app paywall. So they could only do this by making it available to everyone including free users. Not sure what got them over the line but its there now.
So makes me wonder if that will go away - or whether they still enforce that requirement on Android, or what.
They can’t see it if you use your fingerprint or FaceID.
That’s actually a security improvement for some threat models.
Just a thought :)