Systemd programming, 30 months later

Posted Sep 28, 2016 8:50 UTC (Wed) by linuxrocks123 (subscriber, #34648)
In reply to: Systemd programming, 30 months later by Cyberax
Parent article: Systemd programming, 30 months later

Tried again just now for about 15 minutes to see if it still sucks. Just for fun, since I don't use it anymore. I used the same machine for my NFS server five years ago; nothing has changed but upgrades and the kernel (but the kernel has the nfs module). Should still work, right? Wellll, after fixing about 3 other problems, nfsd still won't start and puts "unable to set any sockets for nfsd" in syslog. So yeah, still sucks.

I wasn't doing something wrong: I eventually got it working. I also got a MythTV frontend to work through NFS, but the bandwidth efficiency was worse than HTTP, so I shut that down, too. Yes, I played with the NFS block size first, tried UDP versus TCP, all that.

I remember NFS as being more hellishly frustrating to configure than just about any other standard UNIX daemon I have encountered, which is quite a few. Apparently, it still is.

Systemd programming, 30 months later

Posted Sep 28, 2016 9:56 UTC (Wed) by ballombe (subscriber, #9523) [Link] (3 responses)

I used extensively NFS 15 years ago and it was very easy to setup: just add
a line it fstab on the client and one on the server in exports.

One the other hand, why did you expect NFS to be more efficient than HTTP ?

Systemd programming, 30 months later

Posted Sep 29, 2016 17:23 UTC (Thu) by linuxrocks123 (subscriber, #34648) [Link] (2 responses)

> why did you expect NFS to be more efficient than HTTP ?

I didn't necessarily expect it to be better, but I certainly didn't expect it to be worse, especially given that HTTP is well-known for having relatively high overhead for file transfers.

I forget why I was trying NFS for MythTV in the first place; probably something about the frontend doubling as a slave backend or something.

Systemd programming, 30 months later

Posted Sep 29, 2016 17:43 UTC (Thu) by alankila (guest, #47141) [Link] (1 responses)

Why on Earth would http have high overhead for file transfers? In the simplest case you do a single request for the contents of the whole file. The headers are probably somewhere in order of 100 bytes in both directions, and all the rest is the file's binary data, unmangled.

Systemd programming, 30 months later

Posted Sep 29, 2016 21:30 UTC (Thu) by bronson (subscriber, #4806) [Link]

Presumably when sending tons of small files without keepalive (like, if you're invoking curl for each file). Agreed, pretty obscure, and easily fixed with a tarpipe.