PR#841: kojid: make install timeout of imagefactory conf configurable - koji

koji

#841 kojid: make install timeout of imagefactory conf configurable

Merged 5 years ago by mikem. Opened 6 years ago by julian8628.

julian8628/koji issue/837 into master

change default value of oz_install_timeout to 0

Yuming Zhu • 5 years ago

a49f3df

add explanation in image_build.rst

Yuming Zhu • 5 years ago

c3ef74f

kojid: make install timeout of imagefactory conf configurable

Yuming Zhu • 5 years ago

6b05300

builder/kojid

file modified

+5 -4

		`@@ -3519,8 +3519,8 @@`
		`#IF specific`
		`'imgdir': os.path.join(self.workdir, 'scratch_images'),`
		`'tmpdir': os.path.join(self.workdir, 'oz-tmp'),`
		`- 'verbose' : True,`
		`- 'timeout': 7200,`
		`+ 'verbose': True,`
		`+ 'timeout': self.options.oz_install_timeout or None,`
		`'output': 'log',`
		`'raw': False,`
		`'debug': True,`
		`@@ -5666,7 +5666,7 @@`
		`'max_retries': 120,`
		`'offline_retry': True,`
		`'offline_retry_interval': 120,`
		`- 'keepalive' : True,`
		`+ 'keepalive': True,`
		`'timeout' : None,`
		`'no_ssl_verify' : False,`
		`'use_fast_upload': True,`
		`@@ -5681,6 +5681,7 @@`
		`'resolver-status.properties *.lastUpdated',`
		`'failed_buildroot_lifetime' : 3600 * 4,`
		`'rpmbuild_timeout' : 3600 * 24,`
		`+ 'oz_install_timeout': 0,`
		`'cert': None,`
		`'ca': '', # FIXME: Unused, remove in next major release`
		`'serverca': None}`
		`@@ -5688,7 +5689,7 @@`
		`for name, value in config.items('kojid'):`
		`if name in ['sleeptime', 'maxjobs', 'minspace', 'retry_interval',`
		`'max_retries', 'offline_retry_interval', 'failed_buildroot_lifetime',`
		`- 'timeout', 'rpmbuild_timeout',]:`
		`+ 'timeout', 'rpmbuild_timeout', 'oz_install_timeout',]:`
		`try:`
		`defaults[name] = int(value)`
		`except ValueError:`

builder/kojid.conf

file modified

		`@@ -38,6 +38,11 @@`
		`; Timeout for build duration (24 hours)`
		`; rpmbuild_timeout=86400`

		`+ ; Install timeout(seconds) for image build`
		`+ ; Default value is 0, which means using the number in /etc/oz/oz.cfg,`
		`+ ; supported since oz-0.16.0`
		`+ ; oz_install_timeout=7200`
		`+`
		`; The URL for the xmlrpc server`
		`server=http://hub.example.com/kojihub`

docs/source/image_build.rst

file modified

		`@@ -660,6 +660,11 @@`
		`#. python-psphere => 0.5`
		`#. VMDKStream => 0.2`
		`#. pykickstart`
		+ #. Edit ``/etc/kojid/kojid.conf``, and set an second value, eg: 7200 for
		+ ``oz_install_timeout``. It's a timeout waiting guest installing. Default
		+ value is 0, that means oz will use its default value. Since ``oz-0.16.0``,
		+ it can be configured in ``/etc/oz/oz.cfg`` as ``install`` in ``[timeouts]``
		`+ section.`
		#. Edit ``/etc/oz/oz.cfg``, and set the memory value in the ``[libvirt]``
		section to at least 2048. Set ``safe_generation`` under ``[icicle]`` to yes.
		#. Run: ``mkdir -p ~root/.psphere/templates``, and then copy the following

julian8628 commented 6 years ago

related PR #837
fixes #842

make this value configurable,

if it's not specified in kojid.conf, passing None to imagefactory, who will tell oz to use default timeout which can be specified in /etc/oz/oz.cfg since oz-0.16.0

mikem commented 6 years ago

This defaults to None, where before we had a fixed value of 7200.

When options.oz_install_timeout is None, we pass that along, which leads me to wonder:

is there a difference between passing timeout: None and leaving it out?
or for that matter, does it have a different meaning from a timeout of 0?
is the None value useful? If so, should we provide a way for the user to specify it explicitly?

Edited 6 years ago by mikem

julian8628 commented 6 years ago

is there a difference between passing timeout: None and leaving it out?

If moving 'timeout' key from imagefactory configuration dict, imagefactory will raise KeyError

or for that matter, does it have a different meaning from a timeout of 0?

Aha, 0 is actually ZERO on oz, it won't wait and throw timeout error immediately

is the None value useful? If so, should we provide a way for the user to specify it explicitly?

None value tells oz to use the number in /etc/oz/oz.cfg.
I added some comments in default kojid.conf. I'm going to add notes in docs.

Edited 6 years ago by julian8628

1 new commit added

add explanation in image_build.rst

6 years ago

julian8628 commented 6 years ago

added explanation in image_build.rst

sinnykumari commented 6 years ago

:+1: Thank you @julian8628 for working on this patch.

sinnykumari commented 6 years ago

We need this patch in F27 koji build in order to configure timeout to fix cloud-base image failure for s390x in F28.

mikem commented 5 years ago

Aha, 0 is actually ZERO on oz, it won't wait and throw timeout error immediately

So presumably no one would ever want to do this.

None value tells oz to use the number in /etc/oz/oz.cfg.
I added some comments in default kojid.conf. I'm going to add notes in docs.

The None value is important, but there is no way explicitly configure this since oz_install_timeout is parsed strictly as an int.

Perhaps we should check for the zero value and pass None in that case instead (unless someone can think of a reason to use a literal timeout of zero).

If we did this, then we could continue to have a default timeout.

rebased onto 6b05300

5 years ago

julian8628 commented 5 years ago

@mikem updated

mikem commented 5 years ago

Commit b64151d fixes this pull-request