Arkaitz Jimenez's questions

My actual problem is that a process I run in an rcfile in docker as docker run container bash --rcfile rcfile remains attached to the bash session by sharing the PGID which means that when I Ctrl-C an unrelated command the process receives a SIGINT

So, inside the rcfile its just (nohup command &), when instead of using the rcfile I call it directly in the shell the PGID is different and Ctrl-C does not affect.

From there I deduced that for some reason, bash does not background processes the same while executing the rcfile.

Wrote a small test without docker where I see the PGID behavior.

RCFILE, sleep and bash share PGID

Cobain ~/tmp/testbash$ cat rcfile 
#!/bin/bash
(sleep 10 &)

Cobain ~/tmp/testbash$ bash --rcfile rcfile
arkaitzj@Cobain:~/tmp/testbash$ ps -o  "%p %r %y %x %c"
  PID  PGID TTY          TIME COMMAND
 2883  2883 pts/0    00:00:06 bash
27911 27911 pts/0    00:00:00 bash
27913 27911 pts/0    00:00:00 sleep
27914 27914 pts/0    00:00:00 ps

Command input in the shell, PGID differs

Cobain ~/tmp/testbash$ bash
Cobain ~/tmp/testbash$ ( sleep 10 &)
Cobain ~/tmp/testbash$  ps -o  "%p %r %y %x %c"
  PID  PGID TTY          TIME COMMAND
 2883  2883 pts/0    00:00:06 bash
27999 27999 pts/0    00:00:00 bash
28127 28126 pts/0    00:00:00 sleep
28132 28132 pts/0    00:00:00 ps

What I don't understand of this bash example though, is that in the rcfile case, when I ctrl-C it does not SIGINT the sleep process, when it is sharing the PGID

I have a script that executes apt-get update on several servers and I have to find out whether there are upgrades available or not in order to schedule apt-get upgrades etc...

Looking at the docs I don't see how I could do this task.

Current environment is several Linux boxes in a couple of logically separated "datacenters" running virtual guests that are created and deleted on demand.
DNS and DHCP run in another box and should serve all the virtual servers.

What I want is to create and remove virtual instances on demand without touching DNS and DHCP server configurations, I've certain freedom to prepare the guests but I'd like to avoid consulting central servers on guest creation.

So, ideally I would like just to create a guest with a hostname "node5.dc1.mydomain.com" and start it, it'd get ip from "mydomain.com" dhcp and register at "mydomain.com" DNS. So that I can just create the instance and ssh into it.

I believe some combination of DHCP and DDNS could work, but I want to ensure that there aren't easier alternatives or confirm that DHCP and DDNS is the way to go.

Although i've read some bad stuff about Dynamic DNS..

I've a planned consecutive updates on a debian package I have for internal use on several servers.

Now, the updates have to happen one after each other at different times and some of the servers might jump several versions so updates can't be assumed to have happened for a given server.

So, I know the basics on debian packaging, but I might be skipping some advanced features that could help me here.
My only idea for now is to have a rather complex preinst and postinst scripts that entirely handle all the range of oldversion-newversion.

Any better idea?

I have an app that uses gconf for configuration and I have notice that some apps install a schema file that lists keys and gives descriptions on them.

Is this mandatory or schema files are optional? It is certainly working without them.
My concerns about the schema is that I don't see how my key schema could be explained.

I store account information on /apps/blah..blah/accounts with one folder per account, but I don't see how could I specify that. /apps/blah..blah/accounts/<account_example>/name is a string owned by blahblah I doubt that could work as there might be several or none accounts under that folder.

I have created a debian package that sets up awstats and the required scripts for a couple of hostnames in a server I have.
My problem is with logrotate, I already update the stats with a cron job I set up on package install but I need to update stats on server logs prerotate, gthe server I'm using, lighttpd, already has an entry on /etc/logrotate.d/.

I guess I could just sed the file and add the prerotate update stats endscript but that wouldn't work later when the package gets uninstalled, and I doubt it'll survive a lighttpd package update.

Any ideas?

I'm running a 3rd party vpn daemon that changes resolv.conf on init and checks for changes continuously.

I installed my own local dns cache that gets properly inside resolv.conf but the vpn daemon puts its resolvers before that and I'm getting huge timeouts.

Any idea on how could I use another config file to specify I want to resolve my own stuff or at least reducing the standard dns timeout(5 secs) without touching resolv.conf

Current setup:

br0|
   |-- tun10 -pipe-tun0(192.240.240.1)
   |-- tun11 -pipe-tun1(192.240.240.2)
   |-- tun12 -pipe-tun2(192.240.240.3)

The pipe program is a custom program that forwards data back2back between two tun interfaces.

The idea is puting 2 programs in .2 and .3 while keeping .1 as the local interface in the current machine. The main problem is that I want to route packets to .2 and to .3 through .1 and br0, but as they are local interfaces, the kernel ignores any routing instruction, it just delivers the packet to the proper interface.

Tried iptables, but the nat table doesn't even see ping packets to those ifaces. A "ping 192.240.240.2" delivers a icmp packet with source and dest .2 to tun1, ideally it should deliver a source .1 dest .2 at tun1 through tun0-br0-tun1

Any hint?

Here the output of some commands:

Output