maas-dhcpd not running due to missing dhcpd.conf

[Natalia Litvinova]

After the initial MAAS installation and some time passing dhcpd.conf dissappears:

cat /var/snap/maas/common/maas/dhcpd.conf
cat: /var/snap/maas/common/maas/dhcpd.conf: No such file or directory

As you can see in the log - the dhcp
service was running before: https://pastebin.canonical.com/p/DvpFqfjXqf/

$ sudo maas status
bind9 RUNNING pid 665699, uptime 19:09:53
dhcpd STOPPED Apr 05 02:27 PM
dhcpd6 STOPPED Not started
http RUNNING pid 623091, uptime 22:05:24
ntp RUNNING pid 622475, uptime 22:05:40
proxy RUNNING pid 631436, uptime 21:59:01
rackd RUNNING pid 621842, uptime 22:05:43
regiond RUNNING pid 621844, uptime 22:05:43
syslog RUNNING pid 622477, uptime 22:05:40

Regiond.log: https://pastebin.canonical.com/p/jT7hqztrGM/
Rackd.log: https://pastebin.canonical.com/p/XWhdRNNsZg/

I've hit this bug 2 times on different stages of the deployment.

MAAS version: 3.3.1-13169-g.94920eb1e

Currently this bug has no workaround on how to make DHCP alive again and affects the ongoing deployment

[Natalia Litvinova (natalytvinova)]

Link to the first sosreport:
https://drive.google.com/file/d/1jMO01A4cMOXcn6DcZjI60PX24ZYBVD1k/view?usp=share_link

[Natalia Litvinova (natalytvinova)]

Link to the second sosreport: https://drive.google.com/file/d/13-jsN2aqw-xG1sfbdMr1AqnFeVnH5OLT/view?usp=share_link

[Natalia Litvinova (natalytvinova)]

subscribed ~field-critical

[Igor Brovtsin (igor-brovtsin)]

Requested debug logs for the issue earlier this week, marking this as incomplete for now.

[Launchpad Janitor (janitor)]

[Expired for MAAS because there has been no activity for 60 days.]

[Baby Philip (babyphilip)]

Issue observed in MAAS snap 3.4.1-14343-g + LXD 5.2x on a physical hardware.

Steps to reproduce:

1. Install MAAS
2. Install LXD
3. Disable dhcp and dns in lxd
4. Enable dhcp for lxd bridge in maas
5. Add LXD to MAAS
6. Deploy new VMs
7. Reboot MAAS server - dhcpd service not running and /var/snap/maas/common/maas/dhcpd.conf missing
8. Enable/Disable dhcp in MAAS Fabric UI, restores the /var/snap/maas/common/maas/dhcpd.conf and service also started

Unable to recreate this in a maas + lxd virtual environment

[Jerzy Husakowski (jhusakowski)]

@babyphilip why is this issue field-critical?

[Jerzy Husakowski (jhusakowski)]

Additional comment - why is LXD installed on the same host as MAAS? This is not a supported production configuration, as both LXD and MAAS will compete to manage DHCP, which requires workarounds. Moreover, deploying VMs on the same host as MAAS can exhaust machine resources and lead to outages.

Since the issue was initially reported a year ago, and it looks like the configuration is not a supported one, this issue should not be tagged as field-critical.

Since a year has passed since the original issue, there's little reason to believe that the same symptom is related to the same issue. If it continues to be a problem, please submit a new issue and include all the required logs to help us diagnose.

[Baby Philip (babyphilip)]

Thanks @jhusakowski, I believe field-critical was subscribed earlier by the original reporter. If this is not a supported production configuration, I will advise customer to segregate LXD and MAAS. I feel it would be great if this is documented as an incompatible/non-supported combination, as recently we have seen few cases with these two together.