3CX installation keeps dropping networked storage

Status
Not open for further replies.

Josh@Stable

Customer
Joined
Jan 19, 2021
Messages
11
Reaction score
1
So I'm running 3CX in production on AWS with their Elastic Filesystem providing me with expandable storage for call recording.

I keep encountering an issue where this mounting is interrupted and then can't be resumed because the EFS mount helper and NFS server appear to have been completely removed from the server.

When this happens I can manually login and reinstall the mount helper at which point recording works again just fine, however I'd like to avoid having to do this manually.

Is there any way to stop these services being removed, or am I missing a better way?

P.s. Archiving calls off to google storage, where they can no longer be played back by a 3CX user, is not a better way
 
Hi Josh,

Indeed if you want your oldest recordings to be playable you need to keep them out of the archive.

However your issue here seems to not be related to 3CX but rather something happening on the OS level, disabling your storage. I think this is more of a "linux on AWS" related question rather than a 3CX question and you should start from there.
 
Ah, I see why you think that.

I missed a portion of my issue which is that it appears whenever 3CX does updates, the installed packages that I need for networked file systems are removed, and I believe this is because the 3CX updater is detecting packages outside of some list and uninstalling them.
 
3CX, when it updates, does not remove any packages not directly related to it.
I'm not sure if any Debian security updates that run, may be affecting this. Again, this process does not uninstall packages, but it will update whatever is flagged as vulnerable, and that will also affect any linked/dependent packages.
 
Going into the apt log, I have proof that the phonesystem is causing the removal of the required packages.
It appears to be very unhappy with the presence of rpcbind, which is required by nfs-common. Any ideas why it's unhappy with this, I've experienced no obvious issues having this package installed while 3CX is in operation

Code:
Start-Date: 2021-02-24  11:25:37
Commandline: apt-get -y install ./build/amazon-efs-utils-1.29.1-1_all.deb
Requested-By: ec2-user (1001)
Install: stunnel4:amd64 (3:5.39-2, automatic), rpcbind:amd64 (0.2.3-0.6, automatic), amazon-efs-utils:amd64 (1.29.1), nfs-common:amd64 (1:1.3.4-2.1+deb9u1, automatic)
End-Date: 2021-02-24  11:25:40

 

Start-Date: 2021-02-24  11:33:42
Commandline: /usr/bin/apt -o Dpkg::Options::=--force-confold -y --force-yes purge rpcbind
Requested-By: phonesystem (999)
Purge: rpcbind:amd64 (0.2.3-0.6), amazon-efs-utils:amd64 (1.29.1), nfs-common:amd64 (1:1.3.4-2.1+deb9u1)
End-Date: 2021-02-24  11:33:45
 
I think I see what is happening.
3CX doesn't like rpcbind, as it is in the category of apps that shouldn't be on the same server as 3CX (it's very very similar to a DNS Server).
Now because the amazon-efs-utils depends on rpcbind, that is removed as well in the process.

If the Amazon Helper app depends on rpcbind, and there is not other way to mount it, then I'm afraid there is not a lot that can be done.
Isn't there a way you can mount this drive using fstab, or some other built-in method without having to install extra packages?
 
The AWS helper app is just setting up the specifics of a connection via NFSv4 in /etc/fstab. So I already do use fstab. Based on my googling, rpcbind is only needed for v2 or v3 of NFS, however because debian is debian.. their nfs-common package sets you up for both of these old protocols without making it an optional install when I just need NFSv4
 
Something I did come across actually.. which I'm still investigating is that instead of uninstalling RPCbind, 3CX could instead "mask" the rpcbind service.
This is apparently accomplished via:
sudo systemctl mask rpcbind.service
sudo systemctl mask rpcbind.socket
 
Status
Not open for further replies.