Rare project developers

[Alexander Gall <>]

On Fri, 1 Apr 2022 16:23:21 +0200, Alexander Gall <> said:

> Hi csaba
> As you've already seen, I have completed a first attempt at
> integrating the release manager with freerouter
> (https://bitbucket.software.geant.org/projects/RARE/repos/rare-nix/commits/dab3443cf73dd26a554f1a4a94b5f757ca2b452b)
> The service reconfiguration from within the freerouter service is a
> little bit tricky but it should be ok, I think.  Basically, the
> tna-switch-to-generation alias doesn't do the switch immediatley but
> delays it until the next restart of the service.

> Can you please use PRs while we both work on this repo so we don't
> keep overwriting each other's changes?  I don't think we should put
> the GEANT_TESTBED profile into the default configuration since this is
> what ends up on a fresh ONIE installation and it should not be
> specific to our testbed.

> Is the sticky alias guaranteed to be executed before the external
> processes are started? If so, we can remove the default
> /etc/freertr/p4-profile file from the ONIE installer. If not, I guess
> we can still let the alias create the file and accept that there will
> be a couple of failed starts of the bfswd and bffwd processes until
> the file exists.

Just tested it and bfswd gets started before the sticky alias, but
there is only one restart necessary

2022-04-01 14:28:26 info cfgPrcss.doRound:cfgPrcss.java:443 restarting 
process bfswd
2022-04-01 14:28:27 info cfgInit.executeSWcommands:cfgInit.java:734 line 135: 
"alias exec tna-set-profile sticky-param CPE" : Setting bf_router profile to 
"CPE"  Use  reload warm  or  reload process bfswd  reload process bffw
d  to activate      
2022-04-01 14:28:27 info cfgPrcss.doRound:cfgPrcss.java:443 restarting 
process bfswd

I guess that's ok. I'll remove /etc/freertr/p4-profile from the ONIE
image and propagate the default profile from
https://bitbucket.software.geant.org/projects/RARE/repos/rare-nix/browse/default.nix#38
to the initial rtr-sw.txt.

--
Alex

> When changing the profile, it would be enough to restart the bfswd and
> bffwd processes, but this currently fails for bfswd. That's because
> the process has several subprocesses and "reload process ... stop"
> does not kill them.  The top-level script start_bfswd.sh is the leader
> of a new process group, so it would be enough to kill that group.  Can
> you please modify the reload procedure so it either kills the group
> (but only if the process is the group leader) or kills all
> subprocesses explicitly?

> -- 
> Alex