This reverts commit ff6d73b989c57db79bb53e82f814d651945c3ac6.
This brings back zuul inline status which is further optimized in
follow on patches. We do this in phases because it's a lot easier to
review that way.
It also introduces zuul_inline global control variable, so we can land
this turned off, and enable in the debugger to verify it's doing the
correct thing.
Change-Id: If951a4a05a7dfef7a6d0c8e15ed263ff0523165e
This provides the facility to leave tracing in hideci that won't
trigger by default, however if trying to debug bad behavior you can
enter the javascript console, set hide_ci_debug = true.
Change-Id: I42bf4e830256ee64eb726e3a82d805f7ca35e246
This is a small tests server for testing hideci (and other gerrit
static file changes), it does this by creating a local http server,
then proxying and translating pages on the file (changing hard coded
references to zuul.o.o and review.o.o to route through this
file). You'll end up with a localhost:8001 (port is configurable) that
you can browse to and be on live gerrit + local files served.
Login doesn't work, because the openid redirect magic appears to
happen in a way that we can't intercept (boo), but you can still poke
at a lot of the interface pretty seamlessly this way. Makes for much
better testing.
We need to fetch zuul targeted status urls over http as well because
default python 2.7 apparently blows it's stack on zuul's ssl cert.
Change-Id: Ic28729cea9f2d1aea675025e5fef29f2bcc25164
The apply test does some really simple string manipulation to generate
puppet manifests based on our site.pp on the fly. Unfortunately the old
code assumed there would be no blank lines within a node{} definition.
This assumption is breaking as we work to decouple the
openstack_project::server contents from within openstack_project
manifests and move that into site.pp.
This is a relatively simple fix where we csplit nodes based on the
previous /^}$/ which ends the last node{} definition instead of
splitting on blank lines. Added some comments to make what the parsing
does a bit more clear.
Change-Id: I2cc7317ed90423d6f923c4303a76e091a0bc2ebb
Our current puppet run_all.sh script takes almost 45 minutes to run
puppet agent on all of our nodes. We are using the default concurrency
of 5. Our puppet master should be able to handle a bit more than that.
Run the git/gerrit playbook with a concurrency of 10 and everything else
with a concurrency of 20.
Change-Id: Ia09abb6fa8c699e156aed38d86ce6fd193f3a42d
Maintenance there is reportedly complete.
This reverts commit 42c1760933d5c76880f6760625c6f484d01bbf67.
Change-Id: I4db276a67de575b381838ed9a645058d582f7b3d
The latest version of the wildfly puppet module is 0.2.4. Update to
this version since I've tested it and confirmed it's working.
Change-Id: I122224405ad47d82d836ede4676c8c20d6287409
This reverts commit 45328d7582e11f4a2019f7fcf1a428a3e3a45b02.
HPCloud isn't actually done cleaning up things on their end and we are
noticing that nodepool is still having trouble with the API there.
Unfortunately this means that we actually have fewer nodes for jobs than
if we just used rax alone. So switch back to rax alone until HPCloud is
properly ready for us.
Change-Id: I274c1f05849b861e9e08d367d55c5bf45a759392
Unfortunately, when deployed via puppet, this resulted in the
following error in the logs on start, and Gerrit would not start:
Error in custom provider, java.lang.SecurityException: class
"org.bouncycastle.util.io.TeeOutputStream"'s signer information
does not match signer information of other classes in the same
package
This reverts commit 6906f3b94f4e0fb8f47d66325f4d216502512097.
Change-Id: I6992105081faf55a5ebb3417f799a389851040b0
Requested by the Foundation, it's an invite-only mailing list
with private archives, to discuss and coordinate about analyst
relations. The light privacy is required because topics are
discussed under embargo.
Change-Id: Ie8b2b268c7067b8919896cf12e36c9a0e10f1de1
Rather than install any puppet modules from the puppet forge and rely on
puppet module install to deal with dependencies we install everything
from source by hand without implicit dependency resolution. This means
that all new modules must also have all of their transitive dependencies
explicitly listed here.
Note that this completely matches what we have on the existing puppet
master.
The reason we are making this change is the puppet module install
command isn't very good about installing a specific version if a
different version is already present. Instead it wants you to upgrade
things which may update other deps that we do not want to be updated.
Change-Id: I80ca7a69ad8021fbe1572068d81598e5023d4dcc
The newer puppetdb module is dragging in a newer postgresql module
than our openstack_project::slave_db class supports, so we can no
longer build bare-.* nodepool images which use that. Reverting for
now and we can try again taking this situation into account.
This reverts commit e5b4cba36a47864f899b33065ad4b5884a534291.
Change-Id: I9cce98838f8b1bd2b3aab2bd3fa274506632f5ae
We're moving the content of http://ci.openstack.org to
http://docs.openstack.org/infra and will use this to redirect old
deep links once the content is in place. It's safe to land this in
the meantime since it won't be hit until we switch DNS for
ci.openstack.org.
Change-Id: I6c6091e1dd18931211691972c408db8f3bf81719
We have seen what appears to be a rax api endpoint go away without
nodepool noticing. Set a 60 second api timeout on rax nodepool providers
so that the clients will give up after a 60 second period and reconnect.
Reconnecting should get us connected to a happy endpoint socket.
Change-Id: If27fc2e1d45b35bc255daba6bfbce9615dcc3cf6
