Am 01.10.20 um 15:58 schrieb Damien Martins:
Ok Oliver,
I understand your point now: you want to use wildcard related to the dataset name. Hence, we have to validate this is possible... I could not find anything pointing to wildcard usage in define the dataset name, in this documentation: https://xymon.sourceforge.io/xymon/help/howtograph.html
I'm not a C guy, hence I'll consider documentation only: if there is no mention, it does not exist.
You could try to handle manually (awful answer I know) the serial number of your NVMe's ? Or better: contribute to the code (answer from someone lazy who won't)
If it was written in shell code, I probably could contribute, but I don't know anything about C ... The magic lies in xymond/rrd/do_ncv.c (I think)
I ended up setting "GAUGE" which creates all RRDs with this type and then I deleted all RRDs that need to be "DERIVE". Changed the config, restarted server and client and now everything is as it should... It's not ideal, but ok as a workaround.
It's really scary to see that my desktop sending about 12 write commands per second to the SSD's. This results in 2% wear per week. Ok, it's only 256GB SSD and it hosts two operating systems 24/7, but nevertheless its not a good thing. I moved Firefoxs profile entirely to a ramdisk, because sqlite files like history and cookies grate the SSDs ... Maybe running BTRFS in Raid 1 was a wrong decision? Don't really know.
Regards
Oliver