Previously I set vfs.zfs.arc_max="512M" in /boot/loader.conf.local

when I set it to 1024M all seems working really better.

So for zfs-2.2.6 a too small arc is really a bad idea.

So shame on me ...

Henri

On 10/4/24 17:24, Henri Hennebert wrote:
Hello,

Since I upgrade my rockpro64 to

FreeBSD keystone.lab.bel 14.1-STABLE FreeBSD 14.1-STABLE #0 stable/14- n268744-79c34d704f31: Sun Sep 29 13:38:10 CEST 2024 [email protected]:/usr/ obj/usr/src/arm64.aarch64/sys/GENERIC arm64

when I start

cpuset -c -l 4-5 poudriere bulk -j 14aarch64 -z pine64 -f /usr/local/poudriere/ pine64-pkglst

everything is running at least 2 to 3 times slower than under

FreeBSD 14.1-STABLE #0 stable/14-n267830-339b47f01985

Some build repeatedly fail with eg :

pkg-static: Fail to chmod /afm/public/nanumtype1/nanummjbd7.afm:Bad file 
descriptor

previously this fail was rather uncommon.

Moreover at soon as poudriere bulk start top -SH show:

last pid: 79185;  load averages:  3.53,  3.72,  3.85 up 5+01:17:19  17:14:19
502 threads:   13 running, 463 sleeping, 26 waiting
CPU:  0.1% user,  0.2% nice, 38.8% system,  0.1% interrupt, 60.9% idle
Mem: 77M Active, 422M Inact, 4112K Laundry, 1318M Wired, 1987M Free
ARC: 676M Total, 284M MFU, 66M MRU, 11M Anon, 14M Header, 301M Other
      69M Compressed, 294M Uncompressed, 4.25:1 Ratio
Swap: 8192M Total, 35M Used, 8157M Free

   PID USERNAME    PRI NICE   SIZE    RES STATE    C   TIME    WCPU COMMAND
    0 root         -8    -     0B  2432K CPU1     1  96.8H  99.61% kernel{arc_prune}    11 root        187 ki31     0B    96K RUN      0  98.0H  98.76% idle{idle: cpu0}    11 root        187 ki31     0B    96K RUN      1  94.6H  97.63% idle{idle: cpu1}
26388 root        135    4   110M    78M CPU5     5  82:53  94.30% bsdtar
   11 root        187 ki31     0B    96K RUN      3  91.9H  79.92% idle{idle: cpu3}    11 root        187 ki31     0B    96K RUN      2  98.8H  61.87% idle{idle: cpu2}    11 root        187 ki31     0B    96K RUN      5  26.5H  27.09% idle{idle: cpu5}
......

arc_prone running at 99% on one of the CPU all the time, note that 1987M Free.

Does anyone encounter this sort of problems?

Thanks for your time

Henri





Reply via email to