[Labs-l] List of packages that are supported by tool labs

YiFei zhuyifei1999 at gmail.com
Thu Nov 7 12:34:54 UTC 2013


My suggestion: Crontab "dpkg -L" (as root or give a specific user/group to run it) on every node and somehow save stdout (and/or stderr) to NFS. And cron a script that process those. And finally have a web script to let people see them. 

> On Nov 7, 2013, at 8:24 PM, Petr Bena <benapetr at gmail.com> wrote:
> 
> Hi,
> 
> I think we should make a simple list (even autogenerated) of all
> packages that are officialy supported by tool labs and are guaranteed
> to be present on all execution nodes.
> 
> I know that I can open puppet definitions hierarchy for labs, check
> recursively all templates we have from root to execution node and
> check all these lists for my package X, but that is kind of hard-core
> work to do for every single package.
> 
> We can't expect users of tool labs to ever do this to figure out, so
> there needs to be a better list.
> 
> 
> For example right now I need to be sure that package g++, make and
> qt4-qmake are present. On some nodes it may be, on some other nodes it
> may not be. I have root on tool labs and I can directly access every
> server, check dpkg -l there, but regular users do not have this
> ability so how they are supposed to verify that? These packages I
> mentioned, like g++ are very common and probably even included by some
> other puppet template than exec node, so I can't check against puppet
> for exec node if this package is there, but it MAY BE (who knows)
> present in some other template / definition inherited from other
> puppet classes...
> 
> So right now this it's a little bit tricky to figure out if some
> package is available and supported. Even if it is NOW on exec nodes,
> nothing guarantees you that it will be there tommorow, or that it will
> be there on future exec nodes (we have many exec nodes with different
> configuration than other, that already causes lot of troubles, but
> AFAIK I am not supposed to care about that)
> 
> Is there some simple solution for this?
> 
> _______________________________________________
> Labs-l mailing list
> Labs-l at lists.wikimedia.org
> https://lists.wikimedia.org/mailman/listinfo/labs-l



More information about the Labs-l mailing list