Hello Matthias, It seems like you are silencing the symptoms here, but I am perfectly happy with that.
Collectd does not seem to receive a lot of development any more and Adolf has pointed it our that any current RCs are not really a release candidate at all any more. So many plugins are not ported and for a long time we have been thinking about replacing it. I have an implementation on the go which is quite promising, but it is not ready for Core Update 200, yet. I hope it will make it into 201 or 202. That is why I am happy to just leave collectd as it is now and move on. However, we will need to fix this problem in the new implementation that you can find here: https://git.ipfire.org/?p=collecty.git;a=summary When we introduced a lot of the Meltdown/Spectre measures, we usually disable HT on all systems. Nobody has reported since that this might cause any problems in collecting any of the CPU data. As soon as the new solution is available, please make sure that I got it right please. -Michael > On 22 Jan 2026, at 14:04, Matthias Fischer <[email protected]> > wrote: > > On 22.01.2026 12:02, Adolf Belka wrote: >> Hi Matthias, > > Hi Adolf, > > thanks for the feedback! Comments below... > >> On 21/01/2026 21:47, Matthias Fischer wrote: >>> Hi, >>> >>> just in case that this happens to somebody else: >>> >>> Some days ago I made a fresh install of Core 199 with a new machine and >>> rather old cpu (i7-2600). Runs fine and has more power than the old Duo >>> box. Fits my needs. >>> >>> But I couldn't disable HT in BIOS - 'htop' finds eight CPUs. Should be >>> no problem - but it is. >>> >>> The problem came with the 'cpufreq'-plugin, my logs were flooded with >>> warnings: >>> "cpufreq plugin: Reading >>> "/sys/devices/system/cpu/cpu4[5,6,7]/cpufreq/scaling_cur_freq" failed." >>> >>> Because of the deactivation, 'scaling_cur_freq' was empty, there were >>> complains about CPU 4-7 (the deactivated ones) and in no time my log was >>> filled with about 10000 'collectd'-entries. >>> >>> My solution was to change the loglevel of the 'Plugin syslog' in >>> 'collectd.conf' from 'info' to 'err'. >>> >>> Can anyone confirm this behaviour? >> Can't confirm it on my system. I have a new mini with a quad core >> celeron. cpufreq-plugin is working fine. No error messages of any kind >> in the collectd system logs since I did my CU199 update and all graphs >> are displaying fine. > > The graphs are absolutely ok here, too and if I hadn't take a look at > the logs I wouldn't have noticed. The problem with my machine seems to > be that the cpufreq-plugin tries to read four *empty* files (concerning > the deactivated "HT-cores"). Theses files are empty, and the plugin > doesn't like that... > > But as I wrote, setting the syslog-loglevel to 'err' solved it. The > plugin is quiet now and everything else is running fine. ;-) > > Best > Matthias > >> >> Regards, >> >> Adolf. >> >>> Best >>> Matthias >>> >>> >> > >
