Closed XioNoX closed 1 year ago
The issue very likely comes from running out of inodes, which unfortunately results in a rather misleading ‘No space left on device’ error. (We manually adjusted the log message in 0.12. to read ’No space or inodes left on device’.)
Without having tested this, I think you also want to add -o nr_inodes=2M
when mounting the tmpfs. We’ll add this to the incantation mentioned in the manual once we did test.
(To check inode numbers use df -i
)
I'm in the process of documenting this in #890. Still need to test.
FYI we bumped the number of inodes in our Puppet code as suggested and it did the right thing without disrupting Routinator 3000. Thanks!
$ df -i
Filesystem Inodes IUsed IFree IUse% Mounted on
tmpfs 2097152 653093 1444059 32% /var/lib/routinator/repository
$ df -h
Filesystem Size Used Avail Use% Mounted on
tmpfs 4.0G 2.3G 1.8G 56% /var/lib/routinator/repository
Hi,
We're running Routinator 0.11.3 with a 4G tmpfs for the RPKI cache.
Unfortunately, our Routinator systemd service started to alert as it was stuck in a crash loop. The daemon might have been restarted by our automation once (eg. to pick up new library upgrades). And then the following kept happening:
I manually stopped the deamon and cleaned the tmpfs filesystem then restarted Routinator.
It seems to be stable since then.
With the available space on tmpfs I'm a bit surprised of the error
No space left on device
. One hypothesis is that Routinator downloads data first before removing old data during initialization, requiring briefly the double of space/memory.