Search results

Was a leftover `todo!()` entry in the code I somehow missed, should work with the next proxmox-backup-client update (0.8.7-1).

Ich nehme an du benutzt `pct enter`? Denn das Verhalten hier liegt daran in welchem zustand das signal handling ist. `pct` setzt wohl `SIGPIPE` auf ignore, was dazu führt, dass `tr` den pipe fehler beim `write` bekommt, statt vom standard `SIGPIPE` handler terminiert zu werden. (Mit `lxc-attach`...

Are there any lxcfs related syslog entries during that time?

The limits from cgroupv2 definitely work better, but keep in mind that they will only work for block device based storages: LVM, LVM-Thin, iSCSI, RBD, and Directory depending on where the directory lies. Network storages work differently and ZFS is its own magic.

There's a new lxcfs release in the pvetest repository, lxcfs-4.0.3-pve1, can you try and see if it fixes your issue?

So a regular user cannot even do an ls / on your host? lxc needs to open a file handle to / to switch into the guest file system. That's not possible without read access. Do you have a reason for this or did it happen by accident somehow? The regular permissions are 0755 for /.

So there seems to have indeed be a change between lxcfs 3 and 4 that could potentially be the culprit. (A cache for an expensive lookup was not being filled anymore). There's a fix in the stable-4 branch and we'll be shipping it with lxcfs-4.0.2-pve3 soon. I'll let you know when it's ready in...

The main reason the "import" fails is because the whole file system is suddenly under an extra directory. Might be enough to just extract it and re-archive only the rootfs contents. I'd recommend also removing any dev/ contents from the image as those may fail to extract if you're using...

Curious. What do you get from stat / on a shell?

The traces show a lot of processes waiting for a single mutex in lxcfs. Perhaps this can be improved, but at first glance there's not really a "bug" visible. It's just that lxcfs seems to be doing quite a bit of work in a mutually-exclusive code path. However, this isn't new. Are you sure this...

I'm afraid I cannot help with that.

For the most part the same as with ipv4 with some differences. The main sysctl is is in the "all" interface section: net.ipv6.conf.all.forwarding=1, instead of net.ipv4.ip_forward Your provider will likely expect the IP to come from your host's MAC address, in this case enable `proxy_ndp` via a...

You can do regular routing for ipv6 if you use your host as a gateway. Depending on your environment, you'll most likely need to enable NDP-proxying, either manually or with a daemon such as ndppd. Best read up on ndppd, it's way more convenient ;-)

I do not know the details, as this depends heavily on the implementation of resize2fs. It has to copy all the data beyond the targeted size to the front, and in doing so will of course start using more real disk space. In theory it could also deallocate the data it already moved forward, keeping...

It would be nice to know where exactly lxcfs is stuck there. Could you try to get a backtrace while the hangs are happening (you'll need gdb and lxcfs-dbgsym installed)? (Note that If you've been upgrading or downgrading it on the running machine you may need to restart lxcfs (including the...

Right, looks like that's the case. The thing is, these options change a few things around, and depending on what strategy the tools such as htop use to get their information, they may see more or less. Isolating these things properly is next to impossible. We can look into this more soon. (Note...

That functionality is not enabled by default. You need to modify the lxcfs.service to pass `--enable-cfs` # /etc/systemd/system/lxcfs.service.d/override.conf [Service] ExecStart= ExecStart=/usr/bin/lxcfs --enable-cfs /var/lib/lxcfs <rant mode> Code which tries to spawn processes based on the...

Looks like a bug in the openntpd package of debian. The profile doesn't seem to list the files it wants to access. This apparently has happened before [1]. [1] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=799084;msg=5

We currently do not automate this, but we might add support for this in the future. We already create symlinks to `/dev/mapper` (those are the other way round atm.). Adding this would require additionally reading out the uuids during that setup step and adding the links. Come to think of it, we...

Unprivileged containers cannot use `mknod`, you need to bind mount the devices instead.