[dev][oslo] oslo.cache and dogpile 0.7.0+ cache errors
Hi,
Just a heads up that the latest version of dogpile (0.7.0 onwards) have become incompatible with oslo.cache. This is causing a few issues for jobs. It's a little complex due to functional code and many decorated functions.
The error you will see is:
*oslo_cache.**tests.test_**cache.CacheRegi**onTest.**test_function_* *key_generator_* *with_kwargs -------**-------**-------**-------**-------**-------**-------* *-------**-------**-------**-------**------*
*Captured traceback: ~~~~~~~~~~~~~~~~~~~ b'Traceback (most recent call last):' b' File "/tmp/oslo.**cache/oslo_**cache/tests/**test_cache.**py", line 324, in test_function_**key_generator_* *with_kwargs' b' value=self.**test_value)* *' b' File "/tmp/oslo.**cache/.**tox/py37/**lib/python3.**7/site-* *packages/**testtools/**testcase.* *py", line 485, in assertRaises' b' self.assertThat* *(our_callable, matcher)' b' File "/tmp/oslo.**cache/.**tox/py37/* *lib/python3.**7/site-**packages/**testtools/**testcase.*
*py", line 498, in assertThat' b' raise mismatch_error' b'testtools* *.matchers.**_impl.MismatchE**rror: <function CacheRegionTest* *._get_cacheable**_function.**<locals>**.cacheable_**function at 0x7fec3f795400> returned <oslo_cache.**tests.test_**cache.TestProxy**Value object at 0x7fec3f792550>'*
The problem appear since we uncap dogpile.cache on oslo.cache: https://github.com/openstack/oslo.cache/commit/62b53099861134859482656dc92db...
The following unit test fail since we uncap dogpile => https://github.com/openstack/oslo.cache/blob/master/oslo_cache/tests/test_ca...
The problem was introduced by: https://gerrit.sqlalchemy.org/#/c/sqlalchemy/dogpile.cache/+/996/
Your main issue on oslo.cache side is that keyword arguments are tranformed in positionnal arguments when we use dogpile.cache.region.cache_on_arguments.
I've try to revert the changes introduced by the previous dogpile.cache change and everything works fine on the oslo.cache side when changes was reverted (reverted to revision https://github.com/sqlalchemy/dogpile.cache/blob/2762ada1f5e43075494d91c512f... ).
The expected behavior is that dogpile.cache.util.function_key_generator raise a ValueError if **kwargs founds, but our kwargs is empty and our `value=self.test_value was` is recognized as a positionnal argument. Our unit test looking for an assertRaise(ValueError) on cachable decorated function when we pass kwargs but it doesn't happen due to empty kwargs.
For these reasons we guess that is an dogpile.cache issue and not an oslo.cache issue due to the changes introduced by `decorator` module.
The following are related:
- https://github.com/openstack/oslo.cache/blob/master/oslo_cache/tests/test_ca... : unit test where the problem occure - https://review.openstack.org/#/c/638788/ : possible fix but we don't think that is the right way - https://review.openstack.org/#/c/638732/ : possible remove of the unit test who fail
The issue is being tracked in:
https://bugs.launchpad.net/oslo.cache/+bug/1817032
If some dogpile expert can take a look and send feedback on this thread you are welcome.
Thanks,
Copying Mike. More thoughts inline.
On 2/26/19 9:24 AM, Herve Beraud wrote:
Hi,
Just a heads up that the latest version of dogpile (0.7.0 onwards) have become incompatible with oslo.cache. This is causing a few issues for jobs. It's a little complex due to functional code and many decorated functions.
The error you will see is: / /
/oslo_cache.//tests.test_//cache.CacheRegi//onTest.//test_function_//key_generator_//with_kwargs -------//-------//-------//-------//-------//-------//-------//-------//-------//-------//-------//------/
//
/Captured traceback:
b'Traceback (most recent call last):' b' File "/tmp/oslo.//cache/oslo_//cache/tests///test_cache.//py", line 324, in test_function_//key_generator_//with_kwargs' b' value=self.//test_value)//' b' File "/tmp/oslo.//cache/.//tox/py37///lib/python3.//7/site-//packages///testtools///testcase.//py", line 485, in assertRaises' b' self.assertThat//(our_callable, matcher)' b' File "/tmp/oslo.//cache/.//tox/py37///lib/python3.//7/site-//packages///testtools///testcase.//py", line 498, in assertThat' b' raise mismatch_error' b'testtools//.matchers.//_impl.MismatchE//rror: <function CacheRegionTest//._get_cacheable//_function.//<locals>//.cacheable_//function at 0x7fec3f795400> returned <oslo_cache.//tests.test_//cache.TestProxy//Value object at 0x7fec3f792550>'/ The problem appear since we uncap dogpile.cache on oslo.cache: https://github.com/openstack/oslo.cache/commit/62b53099861134859482656dc92db81243b88bd9 The following unit test fail since we uncap dogpile => https://github.com/openstack/oslo.cache/blob/master/oslo_cache/tests/test_cache.py#L318 The problem was introduced by: https://gerrit.sqlalchemy.org/#/c/sqlalchemy/dogpile.cache/+/996/ Your main issue on oslo.cache side is that keyword arguments are tranformed in positionnal arguments when we use dogpile.cache.region.cache_on_arguments. I've try to revert the changes introduced by the previous dogpile.cache change and everything works fine on the oslo.cache side when changes was reverted (reverted to revision https://github.com/sqlalchemy/dogpile.cache/blob/2762ada1f5e43075494d91c512f7c1ec68907258/dogpile/cache/region.py). The expected behavior is that dogpile.cache.util.function_key_generator raise a ValueError if **kwargs founds, but our kwargs is empty and our `value=self.test_value was` is recognized as a positionnal argument. Our unit test looking for an assertRaise(ValueError) on cachable decorated function when we pass kwargs but it doesn't happen due to empty kwargs. For these reasons we guess that is an dogpile.cache issue and not an oslo.cache issue due to the changes introduced by `decorator` module. The following are related: - https://github.com/openstack/oslo.cache/blob/master/oslo_cache/tests/test_cache.py#L318 : unit test where the problem occure - https://review.openstack.org/#/c/638788/ : possible fix but we don't think that is the right way - https://review.openstack.org/#/c/638732/ : possible remove of the unit test who fail
As I noted in the reviews, I don't think this is something we should have been testing in oslo.cache in the first place. The failing test is testing the dogpile interface, not the oslo.cache one. I've seen no evidence that oslo.cache is doing anything wrong here, so our unit tests are clearly testing something that should be out of scope.
And to be clear, I'm not even sure this is a bug in dogpile. It may be a happy side-effect of the decorator change that the regular decorator now works for kwargs too. I don't know dogpile well enough to make a definitive statement on that though. Hence cc'ing Mike. :-)
The issue is being tracked in:
https://bugs.launchpad.net/oslo.cache/+bug/1817032
If some dogpile expert can take a look and send feedback on this thread you are welcome.
Thanks,
-- Hervé Beraud Senior Software Engineer Red Hat - Openstack Oslo irc: hberaud -----BEGIN PGP SIGNATURE-----
wsFcBAABCAAQBQJb4AwCCRAHwXRBNkGNegAALSkQAHrotwCiL3VMwDR0vcja10Q+ Kf31yCutl5bAlS7tOKpPQ9XN4oC0ZSThyNNFVrg8ail0SczHXsC4rOrsPblgGRN+ RQLoCm2eO1AkB0ubCYLaq0XqSaO+Uk81QxAPkyPCEGT6SRxXr2lhADK0T86kBnMP F8RvGolu3EFjlqCVgeOZaR51PqwUlEhZXZuuNKrWZXg/oRiY4811GmnvzmUhgK5G 5+f8mUg74hfjDbR2VhjTeaLKp0PhskjOIKY3vqHXofLuaqFDD+WrAy/NgDGvN22g glGfj472T3xyHnUzM8ILgAGSghfzZF5Skj2qEeci9cB6K3Hm3osj+PbvfsXE/7Kw m/xtm+FjnaywZEv54uCmVIzQsRIm1qJscu20Qw6Q0UiPpDFqD7O6tWSRKdX11UTZ hwVQTMh9AKQDBEh2W9nnFi9kzSSNu4OQ1dRMcYHWfd9BEkccezxHwUM4Xyov5Fe0 qnbfzTB1tYkjU78loMWFaLa00ftSxP/DtQ//iYVyfVNfcCwfDszXLOqlkvGmY1/Y F1ON0ONekDZkGJsDoS6QdiUSn8RZ2mHArGEWMV00EV5DCIbCXRvywXV43ckx8Z+3 B8qUJhBqJ8RS2F+vTs3DTaXqcktgJ4UkhYC2c1gImcPRyGrK9VY0sCT+1iA+wp/O v6rDpkeNksZ9fFSyoY2o =ECSj -----END PGP SIGNATURE-----
FYI dogpile.cache issue was opened: https://github.com/sqlalchemy/dogpile.cache/issues/144
Come with a possible oslo.cache solution that I've introduce there => https://review.openstack.org/#/c/638788/8
Le mar. 26 févr. 2019 à 16:49, Ben Nemec openstack@nemebean.com a écrit :
Copying Mike. More thoughts inline.
On 2/26/19 9:24 AM, Herve Beraud wrote:
Hi,
Just a heads up that the latest version of dogpile (0.7.0 onwards) have become incompatible with oslo.cache. This is causing a few issues for jobs. It's a little complex due to functional code and many decorated functions.
The error you will see is: / /
/oslo_cache.//tests.test_//cache.CacheRegi//onTest.//test_function_//key_generator_//with_kwargs
-------//-------//-------//-------//-------//-------//-------//-------//-------//-------//-------//------/
//
/Captured traceback:
b'Traceback (most recent call last):' b' File "/tmp/oslo.//cache/oslo_//cache/tests///test_cache.//py", line 324, in test_function_//key_generator_//with_kwargs' b' value=self.//test_value)//' b' File"/tmp/oslo.//cache/.//tox/py37///lib/python3.//7/site-//packages///testtools///testcase.//py",
line 485, in assertRaises' b' self.assertThat//(our_callable, matcher)' b' File
"/tmp/oslo.//cache/.//tox/py37///lib/python3.//7/site-//packages///testtools///testcase.//py",
line 498, in assertThat' b' raise mismatch_error' b'testtools//.matchers.//_impl.MismatchE//rror: <function
CacheRegionTest//._get_cacheable//_function.//<locals>//.cacheable_//function
at 0x7fec3f795400> returned <oslo_cache.//tests.test_//cache.TestProxy//Value object at 0x7fec3f792550>'/
The problem appear since we uncap dogpile.cache on oslo.cache:
https://github.com/openstack/oslo.cache/commit/62b53099861134859482656dc92db...
The following unit test fail since we uncap dogpile =>
https://github.com/openstack/oslo.cache/blob/master/oslo_cache/tests/test_ca...
The problem was introduced by: https://gerrit.sqlalchemy.org/#/c/sqlalchemy/dogpile.cache/+/996/
Your main issue on oslo.cache side is that keyword arguments are tranformed in positionnal arguments when we use dogpile.cache.region.cache_on_arguments.
I've try to revert the changes introduced by the previous dogpile.cache change and everything works fine on the oslo.cache side when changes was reverted (reverted to revision
https://github.com/sqlalchemy/dogpile.cache/blob/2762ada1f5e43075494d91c512f... ).
The expected behavior is that dogpile.cache.util.function_key_generator raise a ValueError if **kwargs founds, but our kwargs is empty and our `value=self.test_value was` is recognized as a positionnal argument. Our unit test looking for an assertRaise(ValueError) on cachable decorated function when we pass kwargs but it doesn't happen due to empty kwargs.
For these reasons we guess that is an dogpile.cache issue and not an oslo.cache issue due to the changes introduced by `decorator` module.
The following are related:
https://github.com/openstack/oslo.cache/blob/master/oslo_cache/tests/test_ca...
: unit test where the problem occure
- https://review.openstack.org/#/c/638788/ : possible fix but we don't
think that is the right way
- https://review.openstack.org/#/c/638732/ : possible remove of the
unit
test who fail
As I noted in the reviews, I don't think this is something we should have been testing in oslo.cache in the first place. The failing test is testing the dogpile interface, not the oslo.cache one. I've seen no evidence that oslo.cache is doing anything wrong here, so our unit tests are clearly testing something that should be out of scope.
And to be clear, I'm not even sure this is a bug in dogpile. It may be a happy side-effect of the decorator change that the regular decorator now works for kwargs too. I don't know dogpile well enough to make a definitive statement on that though. Hence cc'ing Mike. :-)
The issue is being tracked in:
https://bugs.launchpad.net/oslo.cache/+bug/1817032
If some dogpile expert can take a look and send feedback on this thread you are welcome.
Thanks,
-- Hervé Beraud Senior Software Engineer Red Hat - Openstack Oslo irc: hberaud -----BEGIN PGP SIGNATURE-----
wsFcBAABCAAQBQJb4AwCCRAHwXRBNkGNegAALSkQAHrotwCiL3VMwDR0vcja10Q+ Kf31yCutl5bAlS7tOKpPQ9XN4oC0ZSThyNNFVrg8ail0SczHXsC4rOrsPblgGRN+ RQLoCm2eO1AkB0ubCYLaq0XqSaO+Uk81QxAPkyPCEGT6SRxXr2lhADK0T86kBnMP F8RvGolu3EFjlqCVgeOZaR51PqwUlEhZXZuuNKrWZXg/oRiY4811GmnvzmUhgK5G 5+f8mUg74hfjDbR2VhjTeaLKp0PhskjOIKY3vqHXofLuaqFDD+WrAy/NgDGvN22g glGfj472T3xyHnUzM8ILgAGSghfzZF5Skj2qEeci9cB6K3Hm3osj+PbvfsXE/7Kw m/xtm+FjnaywZEv54uCmVIzQsRIm1qJscu20Qw6Q0UiPpDFqD7O6tWSRKdX11UTZ hwVQTMh9AKQDBEh2W9nnFi9kzSSNu4OQ1dRMcYHWfd9BEkccezxHwUM4Xyov5Fe0 qnbfzTB1tYkjU78loMWFaLa00ftSxP/DtQ//iYVyfVNfcCwfDszXLOqlkvGmY1/Y F1ON0ONekDZkGJsDoS6QdiUSn8RZ2mHArGEWMV00EV5DCIbCXRvywXV43ckx8Z+3 B8qUJhBqJ8RS2F+vTs3DTaXqcktgJ4UkhYC2c1gImcPRyGrK9VY0sCT+1iA+wp/O v6rDpkeNksZ9fFSyoY2o =ECSj -----END PGP SIGNATURE-----
Submit a patch to dogpile.cache to add some related tests cases:
https://github.com/sqlalchemy/dogpile.cache/pull/145/
Le mar. 26 févr. 2019 à 19:35, Herve Beraud hberaud@redhat.com a écrit :
FYI dogpile.cache issue was opened: https://github.com/sqlalchemy/dogpile.cache/issues/144
Come with a possible oslo.cache solution that I've introduce there => https://review.openstack.org/#/c/638788/8
Le mar. 26 févr. 2019 à 16:49, Ben Nemec openstack@nemebean.com a écrit :
Copying Mike. More thoughts inline.
On 2/26/19 9:24 AM, Herve Beraud wrote:
Hi,
Just a heads up that the latest version of dogpile (0.7.0 onwards) have become incompatible with oslo.cache. This is causing a few issues for jobs. It's a little complex due to functional code and many decorated functions.
The error you will see is: / /
/oslo_cache.//tests.test_//cache.CacheRegi//onTest.//test_function_//key_generator_//with_kwargs
-------//-------//-------//-------//-------//-------//-------//-------//-------//-------//-------//------/
//
/Captured traceback:
b'Traceback (most recent call last):' b' File "/tmp/oslo.//cache/oslo_//cache/tests///test_cache.//py", line 324, in test_function_//key_generator_//with_kwargs' b' value=self.//test_value)//' b' File"/tmp/oslo.//cache/.//tox/py37///lib/python3.//7/site-//packages///testtools///testcase.//py",
line 485, in assertRaises' b' self.assertThat//(our_callable, matcher)' b' File
"/tmp/oslo.//cache/.//tox/py37///lib/python3.//7/site-//packages///testtools///testcase.//py",
line 498, in assertThat' b' raise mismatch_error' b'testtools//.matchers.//_impl.MismatchE//rror: <function
CacheRegionTest//._get_cacheable//_function.//<locals>//.cacheable_//function
at 0x7fec3f795400> returned <oslo_cache.//tests.test_//cache.TestProxy//Value object at 0x7fec3f792550>'/
The problem appear since we uncap dogpile.cache on oslo.cache:
https://github.com/openstack/oslo.cache/commit/62b53099861134859482656dc92db...
The following unit test fail since we uncap dogpile =>
https://github.com/openstack/oslo.cache/blob/master/oslo_cache/tests/test_ca...
The problem was introduced by: https://gerrit.sqlalchemy.org/#/c/sqlalchemy/dogpile.cache/+/996/
Your main issue on oslo.cache side is that keyword arguments are tranformed in positionnal arguments when we use dogpile.cache.region.cache_on_arguments.
I've try to revert the changes introduced by the previous dogpile.cache change and everything works fine on the oslo.cache side when changes
was
reverted (reverted to revision
https://github.com/sqlalchemy/dogpile.cache/blob/2762ada1f5e43075494d91c512f... ).
The expected behavior is that dogpile.cache.util.function_key_generator raise a ValueError if **kwargs founds, but our kwargs is empty and our `value=self.test_value was` is recognized as a positionnal argument. Our unit test looking for an assertRaise(ValueError) on cachable decorated function when we pass kwargs but it doesn't happen due to empty kwargs.
For these reasons we guess that is an dogpile.cache issue and not an oslo.cache issue due to the changes introduced by `decorator` module.
The following are related:
https://github.com/openstack/oslo.cache/blob/master/oslo_cache/tests/test_ca...
: unit test where the problem occure
- https://review.openstack.org/#/c/638788/ : possible fix but we don't
think that is the right way
- https://review.openstack.org/#/c/638732/ : possible remove of the
unit
test who fail
As I noted in the reviews, I don't think this is something we should have been testing in oslo.cache in the first place. The failing test is testing the dogpile interface, not the oslo.cache one. I've seen no evidence that oslo.cache is doing anything wrong here, so our unit tests are clearly testing something that should be out of scope.
And to be clear, I'm not even sure this is a bug in dogpile. It may be a happy side-effect of the decorator change that the regular decorator now works for kwargs too. I don't know dogpile well enough to make a definitive statement on that though. Hence cc'ing Mike. :-)
The issue is being tracked in:
https://bugs.launchpad.net/oslo.cache/+bug/1817032
If some dogpile expert can take a look and send feedback on this thread you are welcome.
Thanks,
-- Hervé Beraud Senior Software Engineer Red Hat - Openstack Oslo irc: hberaud -----BEGIN PGP SIGNATURE-----
wsFcBAABCAAQBQJb4AwCCRAHwXRBNkGNegAALSkQAHrotwCiL3VMwDR0vcja10Q+ Kf31yCutl5bAlS7tOKpPQ9XN4oC0ZSThyNNFVrg8ail0SczHXsC4rOrsPblgGRN+ RQLoCm2eO1AkB0ubCYLaq0XqSaO+Uk81QxAPkyPCEGT6SRxXr2lhADK0T86kBnMP F8RvGolu3EFjlqCVgeOZaR51PqwUlEhZXZuuNKrWZXg/oRiY4811GmnvzmUhgK5G 5+f8mUg74hfjDbR2VhjTeaLKp0PhskjOIKY3vqHXofLuaqFDD+WrAy/NgDGvN22g glGfj472T3xyHnUzM8ILgAGSghfzZF5Skj2qEeci9cB6K3Hm3osj+PbvfsXE/7Kw m/xtm+FjnaywZEv54uCmVIzQsRIm1qJscu20Qw6Q0UiPpDFqD7O6tWSRKdX11UTZ hwVQTMh9AKQDBEh2W9nnFi9kzSSNu4OQ1dRMcYHWfd9BEkccezxHwUM4Xyov5Fe0 qnbfzTB1tYkjU78loMWFaLa00ftSxP/DtQ//iYVyfVNfcCwfDszXLOqlkvGmY1/Y F1ON0ONekDZkGJsDoS6QdiUSn8RZ2mHArGEWMV00EV5DCIbCXRvywXV43ckx8Z+3 B8qUJhBqJ8RS2F+vTs3DTaXqcktgJ4UkhYC2c1gImcPRyGrK9VY0sCT+1iA+wp/O v6rDpkeNksZ9fFSyoY2o =ECSj -----END PGP SIGNATURE-----
-- Hervé Beraud Senior Software Engineer Red Hat - Openstack Oslo irc: hberaud -----BEGIN PGP SIGNATURE-----
wsFcBAABCAAQBQJb4AwCCRAHwXRBNkGNegAALSkQAHrotwCiL3VMwDR0vcja10Q+ Kf31yCutl5bAlS7tOKpPQ9XN4oC0ZSThyNNFVrg8ail0SczHXsC4rOrsPblgGRN+ RQLoCm2eO1AkB0ubCYLaq0XqSaO+Uk81QxAPkyPCEGT6SRxXr2lhADK0T86kBnMP F8RvGolu3EFjlqCVgeOZaR51PqwUlEhZXZuuNKrWZXg/oRiY4811GmnvzmUhgK5G 5+f8mUg74hfjDbR2VhjTeaLKp0PhskjOIKY3vqHXofLuaqFDD+WrAy/NgDGvN22g glGfj472T3xyHnUzM8ILgAGSghfzZF5Skj2qEeci9cB6K3Hm3osj+PbvfsXE/7Kw m/xtm+FjnaywZEv54uCmVIzQsRIm1qJscu20Qw6Q0UiPpDFqD7O6tWSRKdX11UTZ hwVQTMh9AKQDBEh2W9nnFi9kzSSNu4OQ1dRMcYHWfd9BEkccezxHwUM4Xyov5Fe0 qnbfzTB1tYkjU78loMWFaLa00ftSxP/DtQ//iYVyfVNfcCwfDszXLOqlkvGmY1/Y F1ON0ONekDZkGJsDoS6QdiUSn8RZ2mHArGEWMV00EV5DCIbCXRvywXV43ckx8Z+3 B8qUJhBqJ8RS2F+vTs3DTaXqcktgJ4UkhYC2c1gImcPRyGrK9VY0sCT+1iA+wp/O v6rDpkeNksZ9fFSyoY2o =ECSj -----END PGP SIGNATURE-----
To close the loop on this, we just merged a unit test fix that unblocks oslo.cache ci. We'll continue to work on sorting out where these tests should live as a followup.
On 2/26/19 1:40 PM, Herve Beraud wrote:
Submit a patch to dogpile.cache to add some related tests cases:
https://github.com/sqlalchemy/dogpile.cache/pull/145/
Le mar. 26 févr. 2019 à 19:35, Herve Beraud <hberaud@redhat.com mailto:hberaud@redhat.com> a écrit :
FYI dogpile.cache issue was opened: https://github.com/sqlalchemy/dogpile.cache/issues/144 Come with a possible oslo.cache solution that I've introduce there => https://review.openstack.org/#/c/638788/8 Le mar. 26 févr. 2019 à 16:49, Ben Nemec <openstack@nemebean.com <mailto:openstack@nemebean.com>> a écrit : Copying Mike. More thoughts inline. On 2/26/19 9:24 AM, Herve Beraud wrote: > Hi, > > Just a heads up that the latest version of dogpile (0.7.0 onwards) > have become incompatible with oslo.cache. This is causing a few > issues for jobs. It's a little complex due to functional code and many > decorated functions. > > The error you will see is: > / > / > > /oslo_cache.//tests.test_//cache.CacheRegi//onTest.//test_function_//key_generator_//with_kwargs > -------//-------//-------//-------//-------//-------//-------//-------//-------//-------//-------//------/ > > // > > /Captured traceback: > ~~~~~~~~~~~~~~~~~~~ > b'Traceback (most recent call last):' > b' File "/tmp/oslo.//cache/oslo_//cache/tests///test_cache.//py", > line 324, in test_function_//key_generator_//with_kwargs' > b' value=self.//test_value)//' > b' File > "/tmp/oslo.//cache/.//tox/py37///lib/python3.//7/site-//packages///testtools///testcase.//py", > line 485, in assertRaises' > b' self.assertThat//(our_callable, matcher)' > b' File > "/tmp/oslo.//cache/.//tox/py37///lib/python3.//7/site-//packages///testtools///testcase.//py", > line 498, in assertThat' > b' raise mismatch_error' > b'testtools//.matchers.//_impl.MismatchE//rror: <function > CacheRegionTest//._get_cacheable//_function.//<locals>//.cacheable_//function > at 0x7fec3f795400> returned > <oslo_cache.//tests.test_//cache.TestProxy//Value object at > 0x7fec3f792550>'/ > > > The problem appear since we uncap dogpile.cache on oslo.cache: > https://github.com/openstack/oslo.cache/commit/62b53099861134859482656dc92db81243b88bd9 > > The following unit test fail since we uncap dogpile => > https://github.com/openstack/oslo.cache/blob/master/oslo_cache/tests/test_cache.py#L318 > > The problem was introduced by: > https://gerrit.sqlalchemy.org/#/c/sqlalchemy/dogpile.cache/+/996/ > > Your main issue on oslo.cache side is that keyword arguments are > tranformed in positionnal arguments when we use > dogpile.cache.region.cache_on_arguments. > > I've try to revert the changes introduced by the previous dogpile.cache > change and everything works fine on the oslo.cache side when changes was > reverted (reverted to revision > https://github.com/sqlalchemy/dogpile.cache/blob/2762ada1f5e43075494d91c512f7c1ec68907258/dogpile/cache/region.py). > > The expected behavior is that dogpile.cache.util.function_key_generator > raise a ValueError if **kwargs founds, but our kwargs is empty and our > `value=self.test_value was` is recognized as a positionnal argument. > Our unit test looking for an assertRaise(ValueError) on cachable > decorated function when we pass kwargs but it doesn't happen due to > empty kwargs. > > For these reasons we guess that is an dogpile.cache issue and not an > oslo.cache issue due to the changes introduced by `decorator` module. > > The following are related: > > - > https://github.com/openstack/oslo.cache/blob/master/oslo_cache/tests/test_cache.py#L318 > : unit test where the problem occure > - https://review.openstack.org/#/c/638788/ : possible fix but we don't > think that is the right way > - https://review.openstack.org/#/c/638732/ : possible remove of the unit > test who fail As I noted in the reviews, I don't think this is something we should have been testing in oslo.cache in the first place. The failing test is testing the dogpile interface, not the oslo.cache one. I've seen no evidence that oslo.cache is doing anything wrong here, so our unit tests are clearly testing something that should be out of scope. And to be clear, I'm not even sure this is a bug in dogpile. It may be a happy side-effect of the decorator change that the regular decorator now works for kwargs too. I don't know dogpile well enough to make a definitive statement on that though. Hence cc'ing Mike. :-) > > The issue is being tracked in: > > https://bugs.launchpad.net/oslo.cache/+bug/1817032 > > If some dogpile expert can take a look and send feedback on this thread > you are welcome. > > Thanks, > > -- > Hervé Beraud > Senior Software Engineer > Red Hat - Openstack Oslo > irc: hberaud > -----BEGIN PGP SIGNATURE----- > > wsFcBAABCAAQBQJb4AwCCRAHwXRBNkGNegAALSkQAHrotwCiL3VMwDR0vcja10Q+ > Kf31yCutl5bAlS7tOKpPQ9XN4oC0ZSThyNNFVrg8ail0SczHXsC4rOrsPblgGRN+ > RQLoCm2eO1AkB0ubCYLaq0XqSaO+Uk81QxAPkyPCEGT6SRxXr2lhADK0T86kBnMP > F8RvGolu3EFjlqCVgeOZaR51PqwUlEhZXZuuNKrWZXg/oRiY4811GmnvzmUhgK5G > 5+f8mUg74hfjDbR2VhjTeaLKp0PhskjOIKY3vqHXofLuaqFDD+WrAy/NgDGvN22g > glGfj472T3xyHnUzM8ILgAGSghfzZF5Skj2qEeci9cB6K3Hm3osj+PbvfsXE/7Kw > m/xtm+FjnaywZEv54uCmVIzQsRIm1qJscu20Qw6Q0UiPpDFqD7O6tWSRKdX11UTZ > hwVQTMh9AKQDBEh2W9nnFi9kzSSNu4OQ1dRMcYHWfd9BEkccezxHwUM4Xyov5Fe0 > qnbfzTB1tYkjU78loMWFaLa00ftSxP/DtQ//iYVyfVNfcCwfDszXLOqlkvGmY1/Y > F1ON0ONekDZkGJsDoS6QdiUSn8RZ2mHArGEWMV00EV5DCIbCXRvywXV43ckx8Z+3 > B8qUJhBqJ8RS2F+vTs3DTaXqcktgJ4UkhYC2c1gImcPRyGrK9VY0sCT+1iA+wp/O > v6rDpkeNksZ9fFSyoY2o > =ECSj > -----END PGP SIGNATURE----- > -- Hervé Beraud Senior Software Engineer Red Hat - Openstack Oslo irc: hberaud -----BEGIN PGP SIGNATURE----- wsFcBAABCAAQBQJb4AwCCRAHwXRBNkGNegAALSkQAHrotwCiL3VMwDR0vcja10Q+ Kf31yCutl5bAlS7tOKpPQ9XN4oC0ZSThyNNFVrg8ail0SczHXsC4rOrsPblgGRN+ RQLoCm2eO1AkB0ubCYLaq0XqSaO+Uk81QxAPkyPCEGT6SRxXr2lhADK0T86kBnMP F8RvGolu3EFjlqCVgeOZaR51PqwUlEhZXZuuNKrWZXg/oRiY4811GmnvzmUhgK5G 5+f8mUg74hfjDbR2VhjTeaLKp0PhskjOIKY3vqHXofLuaqFDD+WrAy/NgDGvN22g glGfj472T3xyHnUzM8ILgAGSghfzZF5Skj2qEeci9cB6K3Hm3osj+PbvfsXE/7Kw m/xtm+FjnaywZEv54uCmVIzQsRIm1qJscu20Qw6Q0UiPpDFqD7O6tWSRKdX11UTZ hwVQTMh9AKQDBEh2W9nnFi9kzSSNu4OQ1dRMcYHWfd9BEkccezxHwUM4Xyov5Fe0 qnbfzTB1tYkjU78loMWFaLa00ftSxP/DtQ//iYVyfVNfcCwfDszXLOqlkvGmY1/Y F1ON0ONekDZkGJsDoS6QdiUSn8RZ2mHArGEWMV00EV5DCIbCXRvywXV43ckx8Z+3 B8qUJhBqJ8RS2F+vTs3DTaXqcktgJ4UkhYC2c1gImcPRyGrK9VY0sCT+1iA+wp/O v6rDpkeNksZ9fFSyoY2o =ECSj -----END PGP SIGNATURE------- Hervé Beraud Senior Software Engineer Red Hat - Openstack Oslo irc: hberaud -----BEGIN PGP SIGNATURE-----
wsFcBAABCAAQBQJb4AwCCRAHwXRBNkGNegAALSkQAHrotwCiL3VMwDR0vcja10Q+ Kf31yCutl5bAlS7tOKpPQ9XN4oC0ZSThyNNFVrg8ail0SczHXsC4rOrsPblgGRN+ RQLoCm2eO1AkB0ubCYLaq0XqSaO+Uk81QxAPkyPCEGT6SRxXr2lhADK0T86kBnMP F8RvGolu3EFjlqCVgeOZaR51PqwUlEhZXZuuNKrWZXg/oRiY4811GmnvzmUhgK5G 5+f8mUg74hfjDbR2VhjTeaLKp0PhskjOIKY3vqHXofLuaqFDD+WrAy/NgDGvN22g glGfj472T3xyHnUzM8ILgAGSghfzZF5Skj2qEeci9cB6K3Hm3osj+PbvfsXE/7Kw m/xtm+FjnaywZEv54uCmVIzQsRIm1qJscu20Qw6Q0UiPpDFqD7O6tWSRKdX11UTZ hwVQTMh9AKQDBEh2W9nnFi9kzSSNu4OQ1dRMcYHWfd9BEkccezxHwUM4Xyov5Fe0 qnbfzTB1tYkjU78loMWFaLa00ftSxP/DtQ//iYVyfVNfcCwfDszXLOqlkvGmY1/Y F1ON0ONekDZkGJsDoS6QdiUSn8RZ2mHArGEWMV00EV5DCIbCXRvywXV43ckx8Z+3 B8qUJhBqJ8RS2F+vTs3DTaXqcktgJ4UkhYC2c1gImcPRyGrK9VY0sCT+1iA+wp/O v6rDpkeNksZ9fFSyoY2o =ECSj -----END PGP SIGNATURE-----
On Wed, Feb 27, 2019 at 01:04:38PM -0600, Ben Nemec wrote:
To close the loop on this, we just merged a unit test fix that unblocks oslo.cache ci. We'll continue to work on sorting out where these tests should live as a followup.
I haven't really followed this problem, but just to point out that we added dogpile.cache to zuul's projects and added tests against master to the nodepool "-src" functional tests [1] after the interesting issues we had between it and openstacksdk a while ago.
Thus I'm certain you could get some form of smoke-test job running against dogpile master, and that would respect depends-on for pull requests, etc. This can help avoid fire-drills on release days.
Myself or anyone in #openstack-infra would be happy to help, I'm sure :)
-i
[1] https://git.openstack.org/cgit/openstack-infra/nodepool/tree/devstack/plugin...
On 2/27/19 4:42 PM, Ian Wienand wrote:
On Wed, Feb 27, 2019 at 01:04:38PM -0600, Ben Nemec wrote:
To close the loop on this, we just merged a unit test fix that unblocks oslo.cache ci. We'll continue to work on sorting out where these tests should live as a followup.
I haven't really followed this problem, but just to point out that we added dogpile.cache to zuul's projects and added tests against master to the nodepool "-src" functional tests [1] after the interesting issues we had between it and openstacksdk a while ago.
Thus I'm certain you could get some form of smoke-test job running against dogpile master, and that would respect depends-on for pull requests, etc. This can help avoid fire-drills on release days.
Myself or anyone in #openstack-infra would be happy to help, I'm sure :)
Thanks, will keep that in mind. In this case the problem was that one of our unit tests was actually testing dogpile instead of oslo.cache and it was relying on some semi-broken behavior in dogpile. When dogpile fixed that behavior the test started failing.
So in this case I'm not sure functional tests would have found anything since nothing functional actually broke (in fact, something started working). That said, if someone wanted to add a test against master dogpile I would be in favor of that. Even if it wouldn't have helped in this case, we do have a strong dependency on dogpile.
-i
[1] https://git.openstack.org/cgit/openstack-infra/nodepool/tree/devstack/plugin...
participants (3)
-
Ben Nemec
-
Herve Beraud
-
Ian Wienand