On 2013-12-03 22:58, Aaron Bentley wrote:
> Public bug reported:
>
> Attempting to bootstrap hpcloud on 1.17 is neither succeeding or
> failing, but hanging. For CI, we need operations to either succeed
> or fail. Waiting forever is no good.
...
> Attempting to connect to 10.2.53.125:22 ......Command '('juju',
> 'bootstrap', '-e', 'test-release-hp', '--constraints', 'mem=2G',
> '--show-log')' returned non-zero exit status 2
By my count there are 345 dots there.
- From looking at the loop, it polls 1/second to wait for a DNS name,
and then a 'try to connect with a 5s timeout' and 'sleep 5s' before
trying again. So that could be 10s between dots, which would be ~1hr.
Given we are waiting for an instance to come up, I think our old
timeout was around 10min (which is generous, but not out of scope for
clouds like Azure).
As currently implemented "waitSSH" just uses a raw for{} loop that
waits until signalled, so we probably do want an overall timeout
there. I'll see if we have an easy 'wait no more than 10 minutes'
policy, and then implement it.
John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.13 (Cygwin)
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
On 2013-12-03 22:58, Aaron Bentley wrote:
> Public bug reported:
>
> Attempting to bootstrap hpcloud on 1.17 is neither succeeding or
> failing, but hanging. For CI, we need operations to either succeed
> or fail. Waiting forever is no good.
...
> Attempting to connect to 10.2.53.125:22 ......Command '('juju',
> 'bootstrap', '-e', 'test-release-hp', '--constraints', 'mem=2G',
> '--show-log')' returned non-zero exit status 2
By my count there are 345 dots there.
- From looking at the loop, it polls 1/second to wait for a DNS name,
and then a 'try to connect with a 5s timeout' and 'sleep 5s' before
trying again. So that could be 10s between dots, which would be ~1hr.
Given we are waiting for an instance to come up, I think our old
timeout was around 10min (which is generous, but not out of scope for
clouds like Azure).
As currently implemented "waitSSH" just uses a raw for{} loop that
waits until signalled, so we probably do want an overall timeout
there. I'll see if we have an easy 'wait no more than 10 minutes'
policy, and then implement it.
John
=:->
-----BEGIN PGP SIGNATURE----- www.enigmail. net/
e7g8ACgkQJdeBCY SNAAPltwCgiWOf1 YFtoN0ECtaXBQeS Qf8N FyrMoS/ oS9ki9kuj0
Version: GnuPG v1.4.13 (Cygwin)
Comment: Using GnuPG with Thunderbird - http://
iEYEARECAAYFAlK
FBMAnjKh032jxBG
=sP+0
-----END PGP SIGNATURE-----