-
Notifications
You must be signed in to change notification settings - Fork 43
Intermittent failures on docker pull (block devices not present) #135
Comments
The error message might be a red herring. Here is the output from the daemon logs:
in there is this:
which indicates a time out error on a VsockConnection: So the real reason why the This could potentially be a dup of #124 |
I've tried something slightly different https://github.com/rn/opengcs/tree/dial-timeout-hack I'll try to debug more tomorrow (I'm in the UK and it's evening now) |
I can try out your branch and see why the logs aren't showing up. At a quick glance, it looks like you're missing the commit where the gcs logs are sent to the HCS, so maybe that's the reason if you're trying to find them from docker. |
Using your branch, I definitely see the logs. Here is a truncated snippet from
|
Cool, thanks for checking. So this is definitely the same as #124 then. Do you want me to tidy this up (and roll in the |
Yeah, I agree. I originally had it check etimedout, but I realised it was getting masked. With the virtsock update, I think your change of only retrying on etimedout is right. |
yes, only retrying on ETIMEDOUT, other errors error out properly |
[I think this is a known issue, but could not find an issue tracking it]
docker pull busybox
ordocker pull alpine
fail intermittently on a fresh start (with thec:\lcow
directory empty.This is with:
203a54283e0b0d58c1bc9d8d0f0b4bea8503fe37
dockerd
commit:0dbc182b6
I've seen two types of error messages (see below), both related to block devices for layers not being present.
failed to 'ls /sys/bus/scsi/devices/0:0:0:2/block'
:"mkfs.ext4 -q -E lazy_itable_init=1 -O ^has_journal,sparse_super2,uninit_bg,^resize_inode /dev/sda"
failingThe text was updated successfully, but these errors were encountered: