[SOLVED] Ceph Reef and an Opteron CPU, ceph-mon and OSDs crashing.

[paulsk]

Hello there! My thread is similar to the following one - https://forum.proxmox.com/threads/p...no-longer-working-on-amd-opteron-2427.129613/. I read the previous thread thoroughly and I concur that the hardware in question should be retired. I'm looking for any tips on how to weather the wait for better hardware.

Specs of my node:

CPU: Dual AMD Opteron 2373 EE (8) @ 2.100GHz
Memory: 2007MiB / 15983MiB

I have also tried compiling CEPH from source as previously mentioned, but there was no "reef-stable" branch, so I tried the master one and it failed to build for me.

Here's the output for ceph-mon
https://pastebin.com/SPKpcagq

And here is the OSD output
https://pastebin.com/XaG5s7Jn


Thankfully CEPH keeps chugging along on the other two nodes, just complains about reduced redundancy. Any tips on what could I do to improve things while waiting for new hardware?

 

[Philipp Hufnagl]

Hello

Where did the build fail? What error did you get?

 

[paulsk]

Where did the build fail? What error did you get?

Thanks for replying. Here's the tail end of the output for the make command.
https://pastebin.com/QfS9d4qk

Seeing it complain about the actual code made me feel really out of my depth.
For context: I installed the dependencies using the command shown in the previous thread about "automating dependency install", some ended up missing, so I also ran the dependency install .sh script in the "ceph" folder.

 

[paulsk]

Update: I had the feeling that the build errors may also be related to the outdated CPU. So I tried to build it on a different host today, using the -O1 flag as mentioned in the previous thread. It certainly took a lot of time and RAM, but was a success. Ceph-osd and ceph-mon now start fine and I'll be more at peace while waiting for the replacement.

Note: As the specific changes mentioned for ceph/src/erasure-code/jerasure/CMakeLists.txt were already present in the patches folder, I opted for the more generous application of -O1, in the ceph/debian/rules file.