Re: Debian Buster, bsdtcp, systemd
"Stefan G. Weichinger" <[email protected]>
| Newsgroups | gmane.comp.archivers.amanda.user |
|---|---|
| Message-ID | <[email protected]> |
Am 23.07.21 um 08:32 schrieb Stefan G. Weichinger:
> Am 22.07.21 um 14:13 schrieb Andreas Haumer:
>
>> Right.
>> Amanda client services are started by inetd on Debian Buster.
>> (One could probably write a systemd amanda-client unit quite easily,
>> but so what...)
>
> Hm, in my opinion it would be preferable to get rid of inetd in the long
> run especially when systemd is able to run amandad via socket activation.
>
>> Yes, some of our Amanda clients are Debian Buster based and they seem
>> to behave fine.
>>
>> On the backup server ("hegel") we have disklist entries like this
>> (example for client "kant"):
>>
>> kant.intern.xss.co.at / {
>> user-tar
>> auth "bsdtcp"
>> } -1 eth0
>>
>>
>> As mentioned above on the client ("kant") the amanda services are
>> started via inetd.
>>
>> So in /etc/inetd.conf we have the line:
>>
>> amanda stream tcp nowait backup /usr/lib/amanda/amandad amandad
>> -auth=bsdtcp amdump amindexd amidxtaped
>>
>> (we don't use tcpd)
>>
>>
>> In /etc/amanda-security.conf we have the following lines uncommented:
>>
>> runtar:gnutar_path=/usr/bin/tar
>> tcp_port_range=512,1024
>> udp_port_range=840,860
>>
>> Permissions might be important:
>>
>> root@kant:~# ll /etc/amanda-security.conf
>> -rw-r--r-- 1 root root 2036 Jan 3 2021 /etc/amanda-security.conf
>>
>>
>> Then there is /etc/amandahosts:
>>
>> root@kant:~# ll /etc/amandahosts
>> -rw------- 1 backup backup 55 Jan 3 2021 /etc/amandahosts
>>
>> In /etc/amandahosts we have:
>>
>> localhost backup
>> hegel.intern.xss.co.at amanda amdump
>>
>>
>> Finally open up the relevant port on the firewall, if any.
>>
>>
>> That seems to be all we had to do to get the amanda client
>> up and running on Debian clients (I hope I didn't forget
>> something important...)
>
> Thanks, andreas (greetings from austria to austria ..)
>
> I also run several sites with a setup like this, but most of them
> already use "auth ssh". This seems to be slower, at least in some places.
>
> I am not yet happy with my switch to "auth bsdtcp": I expected it to be
> faster ...