I would not want to propagate unbest-practices, but out of energy saving my 3rd node is turned of most of the time. Having the other nodes in a cluster (instead of having just the two seperate nodes) adds a lot of comfort, in my view.
datacenter cluster
- Thread starter PetrolDriver
- Start date
D
Deleted member 205422
Guest
I just wanted to chime in once again with a small q on this, if the OP allows. As I was troubleshooting I found a post, I believe it's yours @fabian:
https://bugzilla.proxmox.com/show_bug.cgi?id=3672#c11
I am not saying I am digging out the since then fixed bug, but from that comment, can I safely assume that manually disabling
pve-ha-lrm and pve-ha-crm no harm is done whan I know I am not using anything HA? How about the watchdog-mux, can I disable that too? I could not find any resource describing any much how the three interplay.
I would not disable them in general, even if you are not using HA. it can be done to actively disarm HA before doing certain operations, but we are working on exposing that in a more friendly manner as well.
D
Deleted member 205422
Guest
Thank you! For anyone tirelessly digging out more on how this works beyond:
https://pve.proxmox.com/pve-docs/chapter-ha-manager.html#_how_it_works
I can point them to start right in:
https://github.com/ThomasLamprecht/pve-ha-manager/tree/master
Last edited by a moderator:
Regarding:
pve-ha-lrm and pve-ha-crm.Could you please be so kind and elaborate on the why?
D
Deleted member 205422
Guest
(That's a question for @fabian but I did not want to press on as my issue must have been unrelated (to what I thought was originally causing it). And also he was kind enough to reply on my other annoying thread where it was my fault asking for explanations where it was my local config fault. )Regarding:pve-ha-lrmandpve-ha-crm.
Could you please be so kind and elaborate on the why?
I would also be curious about the reason, also why have the
watchdog-mux on if not needed.I should just go read the sources how they all play together, it's quite readable even if you don't do Perl actually.
https://github.com/proxmox/pve-ha-manager/tree/master/src/PVE/Service
https://github.com/proxmox/pve-ha-manager/blob/master/src/PVE/HA/CRM.pm
https://github.com/proxmox/pve-ha-manager/blob/master/src/PVE/HA/LRM.pm
Watchdog is good old C:
https://github.com/proxmox/pve-ha-manager/blob/master/src/watchdog-mux.c
EDIT: If I may to inject one more curious q, what was the reason to develop own thing rather than use e.g. pacemaker?
Last edited by a moderator:
because the next package upgrade would try to start them again, or depending on how you disable them, possibly fail to install.(That's a question for @fabian but I did not want to press on as my issue must have been unrelated (to what I thought was originally causing it). And also he was kind enough to reply on my other annoying thread where it was my fault asking for explanations where it was my local config fault. )
I would also be curious about the reason, also why have thewatchdog-muxon if not needed.
that was before my time - but I assume it was just not a good fit at the time.
Also long before my time, but you can find a short motivation here: https://git.proxmox.com/?p=pve-ha-m...560da85455173299ae4aeb02e56352bed7609;hb=HEAD
And while I haven't looked at it personally, I once was told that pacemaker is highly complex.
D
Deleted member 205422
Guest
I get this hunch it's right there in the open.
21 or even a totally different cluster stack. So, the following core requirements
22 got set out:
23
24 - possibility to run with any distributed key/value store which provides some
25 kind of locking with timeouts (zookeeper, consul, etcd, ..)
26 - self fencing using Linux watchdog device
27 - implemented in Perl, so that we can use PVE framework
28 - only work with simply resources like VMs