QA

Two questions:

• Is there an existing workflow to create backups? Snapshots? rsync? Something else?
• I can't seem to login as a user or root - how does one get SSH access on this machine?

cdywan wrote:

Two questions:

• Is there an existing workflow to create backups? Snapshots? rsync? Something else?

if by "snapshots" you mean btrfs or LVM snapshots that is not possible. Rsync or something else is suggested here.

• I can't seem to login as a user or root - how does one get SSH access on this machine?

you ask an existing user with root access to add your key to /root/.ssh/authorized_keys . We can discuss this together with nsinger in some days.

mkittler wrote:

I'd try this if @nicksinger is available again since I don't know much about the concrete system.

Would you recommend to update to SLE 15 right away?

Yes, I suggest you coordinate with nicksinger. Also maybe he already is in the progress to prepare a complete replacement machine unless I confused something.

And about the backup: Where should I store it? Maybe the new storage server?

No, we have "backup.qa.suse.de" which we can use unless it TBs of data as for openQA where we need a special solution.

I tried the whole day to boot a EFI compatible ISO image over HTTP but failed. Nothing I tried was accepted by the server. No openSUSE ISO, no ipxe-payload, nothing.
I now tried the "CD-ROM Image" option in the BMC. It requires a Windows (SAMBA) share. With a protocol version from "Windows NT". This is absolutely ridiculous but seems to have worked with the following smb.conf on my workstation:

[global]
    workgroup = WORKGROUP
    passdb backend = tdbsam
    map to guest = Bad User
    usershare allow guests = Yes
    log level = 3
    log file = /var/log/samba/%m.log
    min protocol = NT1
    max protocol = SMB3

[boot]
    comment = boot
    path = /home/nsinger/Downloads/opensuse
    public = yes
    read only = no
    force user = nsinger

With this I was able to mount the ISO in the BMC which then created a "Virtual CD drive" on the server. I could boot form it and see grub. Next will be an installation of the basesystem.

Base system is installed now. I will now deploy a basic salt infra based on what we already have in https://gitlab.suse.de/qa-sle/qanet-salt

User creation, ssh key management and postgresql installation is done now

added powerdns packages and some basic postgresql configuration for now. Currently struggling to get salt to create the psql user with the right password hash

last friday I figured out that this is caused by an outdated version of salt in openSUSE. I've opened https://bugzilla.opensuse.org/show_bug.cgi?id=1186500 to address that issue and went with md5 encryption for now (https://gitlab.suse.de/qa-sle/qanet-salt/-/commit/6fe695c82527110acaa61e6f1d4391bdf99943b1)

initial database initialization as well as authoritative and recursive powerdns config was added to salt.
With this the server is now a slave for the current qanet and delivers the same results:

selenium ~ » dig holmes-4.qa.suse.de @qanet.qa.suse.de +short
10.162.2.104
selenium ~ » dig holmes-4.qa.suse.de @qanet2.qa.suse.de +short
10.162.2.104

Next I need to figure out what of the old configuration is still valid:

allow-recursion { localnets; localhost; 10.120.0.40; 10.120.0.41; 10.120.0.44; 10.120.0.45 ; 149.44.176.22; 10.160.0.40; 10.160.0.41; 10.160.0.44; 10.160.0.45; 149.44.176.36; 149.44.176.37; 149.44.176.22; 10.162.64.10; 10.0.0.0/8; };
also-notify { 149.44.160.72; 10.160.0.1; 10.160.2.88; 10.100.2.8; 10.100.2.10; 10.162.0.2; };
allow-transfer { 149.44.160.1; 149.44.160.160; 149.44.160.72; 10.120.0.1; 10.120.0.150; 10.120.2.88; 10.160.0.1; 10.160.0.150; 10.160.2.88; 10.100.2.8; 10.100.2.10; 10.162.0.2; };

When trying to conduct the backup I realized quite slow transfer speeds.

I did a benchmark with

qanetnue:/suse/okurz # dd bs=100M count=20 if=/dev/zero | nc -l 42420

and

backup-vm:/home/okurz # nc qanet.qa 42420 | dd of=/dev/null status=progress

and the result is

116733440 bytes (117 MB, 111 MiB) copied, 1948 s, 59.9 kB/s

so abysmal slow network speed.

-> #107437

With #107437 resolved I can continue. Speed looks much better now.

EDIT: Backup complete.

Now I know why backups fill up our root partition on qanet.qa in the past years. /etc/cron.weekly/removeoldqabackups.sh has:

rm /backups/qa.suse.de_201*.tar.gz-* > /tmp/cronout-$mydate

guess why the problem started in the year 2020 :facepalm: . Fixed by replacing with

find /backups/ -mtime 30 -delete > /tmp/cronout-$mydate

Created a full root partition image backup with command on backup.qa

backup-vm:/home/backup/qanet # nc qanet 42420 | pv > sda2_root-$(date +%F).img

and from qanet

dd bs=1M if=/dev/sda2 | nc backup 42420

Trying zypper dup from a SLE12GM iso image mounted yields

qanetnue:/tmp # zypper dup -r sle12gm
Loading repository data...
Reading installed packages...
Computing distribution upgrade...
29 Problems:
Problem: solvable libgcc_s1-4.8.3+r212056-6.3.x86_64 conflicts with libgcc_s1 provided by itself
Problem: solvable libgcc_s1-32bit-4.8.3+r212056-6.3.x86_64 conflicts with libgcc_s1-32bit provided by itself
Problem: solvable libgfortran3-4.8.3+r212056-6.3.x86_64 conflicts with libgfortran3 provided by itself
Problem: solvable libgomp1-4.8.3+r212056-6.3.x86_64 conflicts with libgomp1 provided by itself
Problem: solvable libquadmath0-4.8.3+r212056-6.3.x86_64 conflicts with libquadmath0 provided by itself
Problem: solvable libstdc++6-4.8.3+r212056-6.3.x86_64 conflicts with libstdc++6 provided by itself
Problem: solvable libstdc++6-32bit-4.8.3+r212056-6.3.x86_64 conflicts with libstdc++6-32bit provided by itself
Problem: solvable libffi4-4.8.3+r212056-6.3.x86_64 conflicts with libffi4 provided by itself
Problem: solvable libffi4-4.8.3+r212056-6.3.x86_64 conflicts with libffi4 provided by itself
Problem: solvable libffi4-32bit-4.8.3+r212056-6.3.x86_64 conflicts with libffi4-32bit provided by itself
Problem: solvable libtsan0-4.8.3+r212056-6.3.x86_64 conflicts with libtsan0 provided by itself
Problem: solvable libtsan0-4.8.3+r212056-6.3.x86_64 conflicts with libtsan0 provided by itself
Problem: solvable libffi4-4.8.3+r212056-6.3.x86_64 conflicts with libffi4 provided by itself
Problem: solvable libffi4-4.8.3+r212056-6.3.x86_64 conflicts with libffi4 provided by itself
Problem: solvable libffi4-4.8.3+r212056-6.3.x86_64 conflicts with libffi4 provided by itself
Problem: solvable libffi4-4.8.3+r212056-6.3.x86_64 conflicts with libffi4 provided by itself
Problem: solvable libffi4-4.8.3+r212056-6.3.x86_64 conflicts with libffi4 provided by itself
Problem: solvable libffi4-4.8.3+r212056-6.3.x86_64 conflicts with libffi4 provided by itself
Problem: solvable libffi4-4.8.3+r212056-6.3.x86_64 conflicts with libffi4 provided by itself
Problem: solvable libffi4-4.8.3+r212056-6.3.x86_64 conflicts with libffi4 provided by itself
Problem: solvable libffi4-4.8.3+r212056-6.3.x86_64 conflicts with libffi4 provided by itself
Problem: solvable libffi4-4.8.3+r212056-6.3.x86_64 conflicts with libffi4 provided by itself
Problem: solvable libffi4-4.8.3+r212056-6.3.x86_64 conflicts with libffi4 provided by itself
Problem: solvable libffi4-4.8.3+r212056-6.3.x86_64 conflicts with libffi4 provided by itself
Problem: solvable libffi4-4.8.3+r212056-6.3.x86_64 conflicts with libffi4 provided by itself
Problem: solvable libffi4-4.8.3+r212056-6.3.x86_64 conflicts with libffi4 provided by itself
Problem: ca-certificates-mozilla-1.97-4.5.noarch requires ca-certificates, but this requirement cannot be provided
Problem: solvable libffi4-4.8.3+r212056-6.3.x86_64 conflicts with libffi4 provided by itself
Problem: solvable libffi4-4.8.3+r212056-6.3.x86_64 conflicts with libffi4 provided by itself

Problem: solvable libgcc_s1-4.8.3+r212056-6.3.x86_64 conflicts with libgcc_s1 provided by itself

so quite problematic.

So trying to find and add update repositories.

zypper ar http://dist.suse.de/updates/repo/\$RCE/SLES11-SP4-LTSS-Updates/sle-11-x86_64/SUSE:Updates:SLE-SERVER:11-SP4-LTSS:x86_64.repo

seems to work, mind the escaped \$RCE. Might be we need the non-LTSS directories in parallel.

EDIT: Added the non-LTSS updates repo:

zypper ar http://dist.suse.de/updates/repo/\$RCE/SLES11-SP4-Updates/sle-11-x86_64/SUSE:Updates:SLE-SERVER:11-SP4:x86_64.repo

Now zypper patch looks sane. Calling it once updated zypper, calling it a second time installs much more. Now zypper patch is clean, so is zypper up. But zypper dup is stuck on conflicts:

3 Problems:
Problem: nothing provides libaudit.so.0 needed by pam-32bit-1.1.5-0.17.2.x86_64
Problem: nothing provides libgdbm.so.3 needed by perl-32bit-5.10.0-64.80.1.x86_64
Problem: nothing provides libaudit.so.0 needed by pam-32bit-1.1.5-0.17.2.x86_64

Likely zypper rm -u pam-32bit could work. We shouldn't need that many 32bit packages if any. Done that.

The following packages are going to be REMOVED:
  ConsoleKit-32bit cryptconfig-32bit pam-32bit pam-modules-32bit pam_mount-32bit samba-32bit samba-winbind-32bit sssd-32bit 

I restarted named and dhcpd and they seem to be still working fine.

nicksinger and me yesterday added the sle11sp4 iso back as repo and then conducted a zypper dup ending up in a consistent state. Then we tried to do an online upgrade with sle12gm repos but ran again into the problems shown in #81192#note-34 . We plan to continue next week Tuesday with an medium based migration.

A problem seems to be the one monitor we have in SRV2 does not support a resolution that is found to be vertical 1050 so soon at boot the monitor does not show anything anymore. Still the system booted and was reachable over ssh. DNS server named and dhcpd are running fine so core services are available. Some failed services, e.g. apache2, all seem to be non-critical so to be cared about later. Trying to add update repos we hit a problem that curl could not initialize. Calling ldd $(which curl) revealed that some /usr/lib/vmware directories were in the list so that sounded fishy. We have renamed that vmware folder with extension .old and curl and zypper were running fine so we could add https://updates.suse.de/SUSE/Updates/SLE-SERVER/12-SP3/x86_64/update/SUSE:Updates:SLE-SERVER:12-SP3:x86_64.repo and https://updates.suse.de/SUSE/Updates/SLE-SERVER/12-SP3-LTSS/x86_64/update/SUSE:Updates:SLE-SERVER:12-SP3-LTSS:x86_64.repo and https://ca.suse.de/ and call zypper dup to bring the system in a proper updated state. After that we rebooted two times to check. Initial grub screen looks weird and no real menu shows up, at least not on VGA, but eventually the system boots fine so good enough.

Next tasks:

• DONE: Check all failed systemd services
• Configure automatic updates and reboots, e.g. use salt same as for backup.qa and alike?
  • Requires prior update to SLE12SP5 (see https://software.opensuse.org/package/salt)
• Upgrade to more recent versions of SLE, then sidegrade to Leap. Or just go to Leap 42.3 now and then upgrade assuming we still find according repos
• Cleanup more old cruft, like apache2, 32bit libraries, etc.
• Consider repartitioning with optional move of / from ext3 to btrfs
• Review all .rpm* files in /etc

I would say next step is that we visit Maxtorhof again and side-grade to Leap and then upgrade all the way to 15.4. As http://download.opensuse.org/distribution/leap/ goes all the way down to 42.3 I see it as easiest to go to Leap first. Bonus points for building a poor-mans-KVM with a raspberryPi, or we just connect the serial port to another machine and plug the power into a remote controlled PDU :) We can use qamaster as serial host. PDUs are already connected but need cable tracing. Further idea a backup qanet as VM on qamaster

serial port is now connected to qamaster and a console is reachable. I also re-plugged on of the Y-power-connectors so all 4 PSUs are now connected to qaps09 - see port documentation in https://racktables.suse.de/index.php?page=object&tab=ports&object_id=1610 or in the webinterface of qaps09

Online migration from 12SP3->12SP4->12SP5 done now. System works fine after a reboot now. According to https://documentation.suse.com/sles/15-SP4/html/SLES-all/cha-upgrade-paths.html#sec-upgrade-paths-supported the upgrade path to SLE15 is only supported offline once again so we might consider going to leap straight away.

Thank you. Next step can be again the "kexec into the downloaded iso" approach and we can dare to do this remotely with serial and remote power control :)

At around 1600 CEST a problem was reported that DNS resolution on grenache-1 does not work.

grenache-1:~ # host openqa.suse.de
;; connection timed out; no servers could be reached

ping -c 1 -4 10.160.0.207, the IPv4 address of OSD, is fine, same as ping -c 1 -6 2620:113:80c0:8080:10:160:0:207. Works now after I restarted named on qanet, not sure why.

Logs of named on qanet:

okurz@qanet:~ 0 (master) $ sudo systemctl status named.service
● named.service - LSB: Domain Name System (DNS) server, named
   Loaded: loaded (/etc/init.d/named; bad; vendor preset: disabled)
   Active: active (exited) since Wed 2022-07-13 15:05:41 CEST; 1h 13min ago
     Docs: man:systemd-sysv-generator(8)
  Process: 1798 ExecStart=/etc/init.d/named start (code=exited, status=0/SUCCESS)
    Tasks: 0 (limit: 512)

Jul 13 15:05:40 qanet named[1875]: automatic empty zone: HOME.ARPA
Jul 13 15:05:40 qanet named[1875]: none:104: 'max-cache-size 90%' - setting to 5356MB (out of 5951MB)
Jul 13 15:05:40 qanet named[1875]: configuring command channel from '/etc/rndc.key'
Jul 13 15:05:40 qanet named[1875]: command channel listening on 127.0.0.1#953
Jul 13 15:05:40 qanet named[1875]: configuring command channel from '/etc/rndc.key'
Jul 13 15:05:40 qanet named[1875]: command channel listening on ::1#953
Jul 13 15:05:41 qanet named[1875]: zone qa.suse.de/IN: cloud2.qa.suse.de/NS 'crowbar.cloud2adm.qa.suse.de' has no SIBLING GLUE address records (A or AAAA)
Jul 13 15:05:41 qanet named[1875]: zone qa.suse.de/IN: cloud3.qa.suse.de/NS 'crowbar.cloud3adm.qa.suse.de' has no SIBLING GLUE address records (A or AAAA)
Jul 13 15:05:41 qanet named[1798]: Starting name server BIND ..done
Jul 13 15:05:41 qanet systemd[1]: Started LSB: Domain Name System (DNS) server, named.

okurz@qanet:~ 3 (master) $ sudo systemctl restart named
okurz@qanet:~ 0 (master) $ sudo journalctl -f -u named
-- Logs begin at Wed 2022-07-13 15:04:39 CEST. --
Jul 13 16:19:31 qanet named[5334]: automatic empty zone: HOME.ARPA
Jul 13 16:19:31 qanet named[5334]: none:104: 'max-cache-size 90%' - setting to 5356MB (out of 5951MB)
Jul 13 16:19:31 qanet named[5334]: configuring command channel from '/etc/rndc.key'
Jul 13 16:19:31 qanet named[5334]: command channel listening on 127.0.0.1#953
Jul 13 16:19:31 qanet named[5334]: configuring command channel from '/etc/rndc.key'
Jul 13 16:19:31 qanet named[5334]: command channel listening on ::1#953
Jul 13 16:19:31 qanet named[5334]: zone qa.suse.de/IN: cloud2.qa.suse.de/NS 'crowbar.cloud2adm.qa.suse.de' has no SIBLING GLUE address records (A or AAAA)
Jul 13 16:19:31 qanet named[5334]: zone qa.suse.de/IN: cloud3.qa.suse.de/NS 'crowbar.cloud3adm.qa.suse.de' has no SIBLING GLUE address records (A or AAAA)
Jul 13 16:19:31 qanet named[5268]: Starting name server BIND - Warning: /var/lib/named/var/run/named/named.pid exists! ..done
Jul 13 16:19:31 qanet systemd[1]: Started LSB: Domain Name System (DNS) server, named.

Handling some restarts:

host=openqa.suse.de failed_since="2022-07-13 13:00" openqa-advanced-retrigger-jobs 
result="result='failed'" host=openqa.suse.de failed_since="2022-07-13 13:00" openqa-advanced-retrigger-jobs

{"result":[{"9116842":9124646}],"test_url":[{"9116842":"\/tests\/9124646"}]}
{"enforceable":1,"errors":["Job 9124626 misses the following mandatory assets: hdd\/SLES-15-x86_64-mru-install-minimal-with-addons-Build20220617-1-Server-DVD-Updates-64bit.qcow2\nEnsure to provide mandatory assets and\/or force retriggering if necessary."],"result":[],"test_url":[]}
{"enforceable":1,"errors":["Job 9124628 misses the following mandatory assets: hdd\/SLES-12-SP4-x86_64-mru-install-minimal-with-addons-Build20220627-1-Server-DVD-Updates-64bit.qcow2\nEnsure to provide mandatory assets and\/or force retriggering if necessary."],"result":[],"test_url":[]}
{"result":[{"9122209":9124648}],"test_url":[{"9122209":"\/tests\/9124648"}]}
{"result":[{"9122212":9124649}],"test_url":[{"9122212":"\/tests\/9124649"}]}
{"result":[{"9122215":9124650}],"test_url":[{"9122215":"\/tests\/9124650"}]}
{"result":[{"9122214":9124651}],"test_url":[{"9122214":"\/tests\/9124651"}]}
{"result":[{"9116915":9124652}],"test_url":[{"9116915":"\/tests\/9124652"}]}
{"result":[{"9116910":9124653}],"test_url":[{"9116910":"\/tests\/9124653"}]}
{"result":[{"9116877":9124654}],"test_url":[{"9116877":"\/tests\/9124654"}]}
{"result":[{"9116913":9124655}],"test_url":[{"9116913":"\/tests\/9124655"}]}
{"result":[{"9116919":9124656}],"test_url":[{"9116919":"\/tests\/9124656"}]}
{"result":[{"9116925":9124657}],"test_url":[{"9116925":"\/tests\/9124657"}]}
{"result":[{"9116928":9124658}],"test_url":[{"9116928":"\/tests\/9124658"}]}
{"result":[{"9116902":9124659}],"test_url":[{"9116902":"\/tests\/9124659"}]}
{"result":[{"9116901":9124660}],"test_url":[{"9116901":"\/tests\/9124660"}]}
{"result":[{"9116914":9124661}],"test_url":[{"9116914":"\/tests\/9124661"}]}
{"result":[{"9116917":9124662}],"test_url":[{"9116917":"\/tests\/9124662"}]}
{"result":[{"9116918":9124663}],"test_url":[{"9116918":"\/tests\/9124663"}]}
{"result":[{"9116926":9124664}],"test_url":[{"9116926":"\/tests\/9124664"}]}
{"result":[{"9116920":9124665}],"test_url":[{"9116920":"\/tests\/9124665"}]}
{"result":[{"9116891":9124666}],"test_url":[{"9116891":"\/tests\/9124666"}]}
{"result":[{"9116922":9124667}],"test_url":[{"9116922":"\/tests\/9124667"}]}
{"result":[{"9116911":9124668}],"test_url":[{"9116911":"\/tests\/9124668"}]}
{"result":[{"9116907":9124669}],"test_url":[{"9116907":"\/tests\/9124669"}]}
{"result":[{"9116921":9124670}],"test_url":[{"9116921":"\/tests\/9124670"}]}
{"result":[{"9116927":9124671}],"test_url":[{"9116927":"\/tests\/9124671"}]}
{"result":[{"9116924":9124672}],"test_url":[{"9116924":"\/tests\/9124672"}]}
{"result":[{"9116895":9124673}],"test_url":[{"9116895":"\/tests\/9124673"}]}
{"result":[{"9116903":9124674}],"test_url":[{"9116903":"\/tests\/9124674"}]}
{"result":[{"9116897":9124675}],"test_url":[{"9116897":"\/tests\/9124675"}]}
{"result":[{"9116896":9124676}],"test_url":[{"9116896":"\/tests\/9124676"}]}
{"result":[{"9116876":9124677}],"test_url":[{"9116876":"\/tests\/9124677"}]}
{"result":[{"9116893":9124678}],"test_url":[{"9116893":"\/tests\/9124678"}]}
{"result":[{"9116909":9124679}],"test_url":[{"9116909":"\/tests\/9124679"}]}
{"result":[{"9116912":9124680}],"test_url":[{"9116912":"\/tests\/9124680"}]}
{"result":[{"9116916":9124681}],"test_url":[{"9116916":"\/tests\/9124681"}]}
{"result":[{"9116929":9124682}],"test_url":[{"9116929":"\/tests\/9124682"}]}
{"result":[{"9116931":9124683}],"test_url":[{"9116931":"\/tests\/9124683"}]}
{"result":[{"9116933":9124684}],"test_url":[{"9116933":"\/tests\/9124684"}]}
{"result":[{"9116934":9124685}],"test_url":[{"9116934":"\/tests\/9124685"}]}
{"result":[{"9116935":9124686}],"test_url":[{"9116935":"\/tests\/9124686"}]}
{"result":[{"9123817":9124687}],"test_url":[{"9123817":"\/tests\/9124687"}]}
{"result":[{"9123818":9124688}],"test_url":[{"9123818":"\/tests\/9124688"}]}
{"result":[{"9124251":9124689}],"test_url":[{"9124251":"\/tests\/9124689"}]}
{"result":[{"9124252":9124690}],"test_url":[{"9124252":"\/tests\/9124690"}]}
{"result":[{"9124253":9124691}],"test_url":[{"9124253":"\/tests\/9124691"}]}
{"result":[{"9124393":9124692}],"test_url":[{"9124393":"\/tests\/9124692"}]}
{"result":[{"9124375":9124693,"9124389":9124694,"9124390":9124695,"9124391":9124696,"9124392":9124697}],"test_url":[{"9124375":"\/tests\/9124693","9124389":"\/tests\/9124694","9124390":"\/tests\/9124695","9124391":"\/tests\/9124696","9124392":"\/tests\/9124697"}]}
{"result":[{"9122213":9124698}],"test_url":[{"9122213":"\/tests\/9124698"}]}
{"result":[{"9124451":9124699}],"test_url":[{"9124451":"\/tests\/9124699"}]}
{"result":[{"9124441":9124700}],"test_url":[{"9124441":"\/tests\/9124700"}]}
{"result":[{"9124442":9124701}],"test_url":[{"9124442":"\/tests\/9124701"}]}
{"result":[{"9124487":9124702}],"test_url":[{"9124487":"\/tests\/9124702"}]}
{"result":[{"9124520":9124703}],"test_url":[{"9124520":"\/tests\/9124703"}]}
{"result":[{"9124536":9124705}],"test_url":[{"9124536":"\/tests\/9124705"}]}
{"result":[{"9124401":9124706}],"test_url":[{"9124401":"\/tests\/9124706"}]}
{"result":[{"9124597":9124707}],"test_url":[{"9124597":"\/tests\/9124707"}]}
{"result":[{"9124380":9124708}],"test_url":[{"9124380":"\/tests\/9124708"}]}
{"result":[{"9124623":9124709}],"test_url":[{"9124623":"\/tests\/9124709"}]}
{"result":[{"9124629":9124710}],"test_url":[{"9124629":"\/tests\/9124710"}]}
{"result":[{"9116923":9124711}],"test_url":[{"9116923":"\/tests\/9124711"}]}

I ran while sleep 10; do date && pgrep -a named; done. Starting

Wed Jul 13 17:41:09 CEST 2022
…
Wed Jul 13 17:51:20 CEST 2022
6269 /usr/sbin/named -t /var/lib/named -u named
Wed Jul 13 17:51:30 CEST 2022
6269 /usr/sbin/named -t /var/lib/named -u named
Wed Jul 13 17:51:41 CEST 2022
7460 /bin/sh /etc/init.d/named stop
Wed Jul 13 17:51:51 CEST 2022
7554 /usr/sbin/named -t /var/lib/named -u named

so something/someone started/stopped/restarted named but a new instance was running. Then later

7554 /usr/sbin/named -t /var/lib/named -u named
Wed Jul 13 18:30:06 CEST 2022
7554 /usr/sbin/named -t /var/lib/named -u named
Wed Jul 13 18:30:16 CEST 2022
Wed Jul 13 18:30:26 CEST 2022
Wed Jul 13 18:30:36 CEST 2022
Wed Jul 13 18:30:46 CEST 2022
Wed Jul 13 18:30:56 CEST 2022
Wed Jul 13 18:31:06 CEST 2022

Maybe a conflict with /etc/init.d. So I did:

mkdir /etc/init.d/old_okurz_20220713
mv /etc/init.d/named /etc/init.d/old_okurz_20220713/

but then I got

qanet:/suse/okurz # systemctl start named
Warning: named.service changed on disk. Run 'systemctl daemon-reload' to reload units.
qanet:/suse/okurz # systemctl daemon-reload
You have new mail in /var/mail/root
qanet:/suse/okurz # systemctl start named
Failed to start named.service: Unit named.service failed to load: No such file or directory.

so the file is actually necessary. Reverted. In journalctl --since=today -u named I saw:

Jul 13 18:30:06 qanet named[7554]: mem.c:906: fatal error:
Jul 13 18:30:06 qanet named[7554]: malloc failed: Cannot allocate memory
Jul 13 18:30:06 qanet named[7554]: exiting (due to fatal error in library)
…
Jul 13 18:30:53 qanet systemd-coredump[8941]: Process 7554 (named) of user 44 dumped core.

                                              Stack trace of thread 7556:
                                              #0  0x00007fc030f390d7 raise (libc.so.6)
                                              #1  0x00007fc030f3a4aa abort (libc.so.6)
                                              #2  0x0000557772ca812f n/a (named)
                                              #3  0x00007fc0330c8fe3 isc_error_fatal (libisc.so.1107)
                                              #4  0x00007fc0330d9293 n/a (libisc.so.1107)
                                              #5  0x00007fc0330d75b9 n/a (libisc.so.1107)
                                              #6  0x00007fc0330d993d isc___mem_allocate (libisc.so.1107)
                                              #7  0x00007fc0330dcbe3 isc___mem_strdup (libisc.so.1107)
                                              #8  0x00007fc033ab08e1 n/a (libdns.so.1110)
                                              #9  0x00007fc033ab3015 dns_resolver_createfetch3 (libdns.so.1110)
                                              #10 0x0000557772caed13 n/a (named)
                                              #11 0x0000557772cbc11b n/a (named)

that can certainly explain the problem of disappearing named. Maybe a memory leak?

EDIT: According to https://flylib.com/books/en/2.684.1/limiting_the_memory_a_name_server_uses.html the configuration option datasize 200M; which we have in /etc/named.conf might be the problem. So I commented that option and restarted the service again. I assume it's enough to rely on https://www.zytrax.com/books/dns/ch7/hkpng.html#max-cache-size. Maybe ages ago "datasize" was by default much lower than 200M.

EDIT: 2022-07-13 19:00Z, named still running so seems to be better.

EDIT: 2022-07-14 07:00Z, same named process still running.

#117043 resolved. With https://gitlab.suse.de/qa-sle/qanet-configs/-/commit/6246fc46224606ba5932f6da6e6d6b87cbc722c5 qanet is still running a very limited dhcp server but forwards DNS to dns1.suse.de, dns1.prg2.suse.org, dns2.suse.de which serves qa.suse.de now from https://gitlab.suse.de/OPS-Service/salt/-/merge_requests/4247, blocked on #134051

https://gitlab.suse.de/OPS-Service/salt/-/merge_requests/4447 to remove DNS entries for decomissioned qanet.