[OmniOS-discuss] LX for OmniOS update

[Jim Klimov]

12 августа 2016 г. 19:14:29 CEST, Dan McDonald <danmcd at omniti.com> пишет:
>Basically, you need to delegate the highres_clock permission to the
>zone.  Look for "proc_clock_highres" in your
>/usr/lib/brand/lx/config.xml file.  If it's not there, add the line per
>the commit I indicated.
>
>Dan

Thanks for thoroughly nudging me - indeed, the privilege was not yet packaged, it seems ;)

So there are new random notes to report ;)

Now a lot more services of our Debian 8-based OS image have started, with notable exceptions being networking and a few services like our webserver whose unit instantly failed with '226/NAMESPACE' (that diagnosis includes 'systemd-networkd.service' and 'systemd-resolved.service' too). I'd guess randomly there is some messup with slices or cgroups, but at least a few services report they belong to a 'system.slice' so at least one works ;)

Part of networking problems may be failure to start iptables. Our units are tweaked to ignore errors at this step, by default this may be fatal.

The ifplugd service says that '...start lo0 (or eth0, or lx0 - whatever - //Jim)...select() failed while waiting for return value: Invalid argument'

The 'systemd-setup-dgram-qlen.service' apparently fails to use 'sysctl -w net.unix.max_dgram_qlen=512' and dies.

Like yesterday (forgot to write though), pasting copied text to system console (login after 'zlogin -C') does not work - it only gets the first character from the buffer, and quick typing also loses some chars (didn't hit that one yesterday, annoys a lot today), but console from 'zlogin' is okay in this regard.

Dmesg and journals for many services are still AWOL.

Some networked services respond on a localhost (e.g. ssh - i can even log in), but our stack that tries using a local MQ bus (on IPC protocol for local speediness, bypassing TCP\IP stack) fails. Maybe i'd try to debug/truss/dtrace that next week, might uncover issues in the project :-D At least error-logging already seems lacking in some of our places ;)

An rsyslog daemon has started this time, so I can see that e.g. PAM complains it can not find kernel RLIMIT for nice, rtprio and UNKNOWN, so it falls back to defaults.

Thanks a lot for doing this,
Jim
--
Typos courtesy of K-9 Mail on my Samsung Android