RancherOS 1.1 kernel panic - not syncing: Attempted to kill init

[vchav73]

It will run fine once or twice then after a reboot I get kernel panics. I'm not able to recover. My provider is Profitbricks. I'm happy to try config changes if anyone has suggestions.

EXTRA_CMDLINE: /init
rancher:
  console: centos
  environment:
    EXTRA_CMDLINE: /init
  network:
    dns:
      nameservers:
      - 8.8.8.8
      - 8.8.4.4
    post_cmds:
    - ip route add default via 10.14.194.15
  services_include:
    kvm-vm-tools: true
  state:
    dev: LABEL=RANCHER_STATE
    wait: true
ssh_authorized_keys:

[root@analytics ~]# dmidecode -s system-product-name
Bochs
[root@analytics ~]# virt-what 
kvm

[Raboo]

I'm trying to ipxe boot RancherOS on a 512MB VPS. And I get the same kernel panic. Why do RancherOS require more than 512MB RAM? Is it possible to build a lightweight version? Or is it lost that microservices can be micro?

[vchav73]

The problem I had turned out to be memory related. In my case I was using a 1 GB server and had to go to 2 GB to avoid the kernel panic.

The problem seems to be cloud vendor specific. 1 GB works with some vendors (Digital Ocean) but not with others (Profitbricks).

[SvenDowideit]

@Raboo the biggest memory problem we have atm, is that we're loading images into the system-docker - this spikes the memory usage on initial boot. The current work to convert to runc and containerd should help significantly..

[Raboo]

@SvenDowideit ok, cool. Is there any issue or something I can subscribe to to follow progress? Or do you have a ETA?

[SvenDowideit]

2079 :)

[EugenMayer]

I have the same issue, but the rancher has 24GB ram, same issue:

Pretty much blocked by this, server does not boot anymore.

This happened suddenly, the server was running fine for months

[EugenMayer]

@SvenDowideit if that is out of any interest ( i am not sure i should create a seperate issue for that ) but my kernel panic was caused by a full disk. What i was doing is

1. noticing disk is full
2. resizing the disk in the HV ( i had auto-resize configured in rancher )
3. rebooting rancher -> kernel panic

Fiddling around and not being able to use the new recovery console ( login data? ) i did

1. boot from an 1.1 rancheros iso
2. used fdisk delete /create /e2fscheck /resize2fs to resize the fs
3. booted and the kernel panic was gone.

So it seems like the kernel panic is (in this case) soley cause by a full disk, which i guess we should avoid, thats a lot of hassle for that, especially if you consider the auto-resize being configured but probably never happened do to the early kernel panic.