[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: debci questions

Hello Paul, thanks for the answers
Hope Debconf will be a good one for Terceiro !

I have investigated a bit further:
- it reads $debci_base_dir/lib/environment.sh so get debci_suite value and set a couple of variables like   debci_autopkgtest_incoming_dir, debci_packages_dir, debci_status_dir which are then used in the following functions.
As such I do believe that it doesn't do a full job regarding all series but do work only on debci_serie= specification.
I then can suggest something if needed.

debci-status allows  -s suite parameter but my experience is the suite info is not used there
/usr/share/debci$ debci status -s testing -l |grep -i pass|wc
   7075   14150  438668
/usr/share/debci$ debci status -s unstable -l |grep -i pass|wc
   7075   14150  438668
So I traced execution and if you specify -s suite parameter we set ENV(debci_suite) to the proper value but then in debci-status line 54 in the FILE.join() we use the Debci.config.packages_dir which is set by default config and not updated by our parameter.

Have you already seen that trace while it is executing ?
Exception `NoMethodError' at /usr/lib/ruby/vendor_ruby/debci/html/status.erb:25 - undefined method `[]' for nil:NilClass
Traceback (most recent call last):
    16: from bin/debci-generate-html:33:in `<main>'
    15: from /usr/lib/ruby/vendor_ruby/debci/html.rb:31:in `status'
    14: from /usr/lib/ruby/vendor_ruby/debci/html.rb:151:in `expand_template
     1: from /usr/lib/ruby/vendor_ruby/debci/html/status.erb:23:in `each'
/usr/lib/ruby/vendor_ruby/debci/html/status.erb:25:in `block (4 levels) in expand_template': undefined method `[]' for nil:NilClass (NoMethodError)

More later

On 7/22/2019 10:02 PM, Paul Gevers wrote:
Hi Thierry,

Terceiro is better at this, but I expect him to be busy with Debconf.

On 22-07-2019 10:27, Thierry Fauck@linux.ibm.com wrote:
- chake/debian-ci-config: Do I understand that master node is dedicated
to that task meaning it is the node running the collector ?
In our setup, yes.

How many network interfaces do you need per node 2 or 3 ? are they
required ?
Per worker you mean? I think 1 is enough. I think you're seeing two in
our configuration because that is how AWS works. The arm64 farm I am
trying to get working only uses one IIAC.

- How do you manage the different suites since the tools like
debci-generate-index only use the suite specified in debci_suite variable ?
Our workers generate one lxc container per suite that is supported. In
our case each worker can run tests for all the supported suites.
debci-generate-index runs on the master. I haven't checked what it would
be doing with the debci_suite variable (as it doesn't make sense to me
there, but maybe I am missing something).

I tried to have - for example one system for testing, one for unstable
and one for stable and I modified debci-generate-index to loop against
all suites defined in debci_suite_list but then I figured out that tools
like debci_status reports only the current suite - is that true ? is the
suite info part of the database ?
I'm pretty sure the suite is part of it yes. I can't comment on the
earlier remarks.

And since that time the status charts are completely strange because -
for example - the  total number of packages in the suite is wrong (it
matches the unstable count I had once) or it varies from very few up to
the total count .... meaning nothing is clean.

So really my question is - (even if I don't put in the equation chake
and do manually because I have very few servers for now) how do you
manage the different suites on the master - or do you have multiple
masters or virtual masters ....
No, just one master to rule them all.

-Thanks in advance for your help clarifying my mind with that great tool
Although I can't answer all your questions, I hope it helps. Don't
hesitate to continue asking.


Thierry Fauck @ fr.ibm.com

Reply to: