On 11/23/2014 05:38 PM, Nathan Scott wrote:
> Hi David,
>
> ----- Original Message -----
>> On 11/19/2014 03:29 AM, Nathan Scott wrote:
>>> OK, good stuff. Do you want to merge the PMDA API changes at this
>>> stage?
>>>
>>> I tried cherry-picking (there's lots of other commits in dsmith/dev):
>>> cec13bfd0297ecc755265ba2db69a86daf32a05c
>>> f2f5a51cde0646fcdf35bc2f60798024c0931c9e
>>> 1fc0bbf9d517810e8512fb3a7775b68fc6f64572
>>> ... but there was a fair few test failures (I haven't dug deeper yet -
>>> they all look like python PMDAs failing to start or exiting early on,
>>> from a quick glance).
>> [...]
>> I ran all the qa tests before and after the cherry pick, and saw passes
>> both ways. The python "simple" pmda seems to work fine also.
>
> Odd. Only thing I can think of is a python version difference - I'm
> using RHEL6 on my primary test machine, which has a relatively old
> python version. I'll have another go and look more deeply into the
> failures.
Hmm, I'll try running this code on RHEL6 (instead of F20) and see what I
get.
>>> Or, do you want to wait for that JSON PMDA to be closer to complete so
>>> we have at least one use case for the API? (either that or some form
>>> of specialised API test case will be needed - a real PMDA is probably
>>> easiest, with its tests covering use of the new API).
>>
>> The latest version of the stap_json PMDA on the dsmith/dev branch, while
>> not finished does use the new refresh_metrics API.
>>
>> I'm flexible on how we proceed here, we can do whatever you think makes
>> the most sense.
>
> We can go either way, just need an appropriate level of testing to go in
> with the API addition. Since there's no rush, and a release in a week -
> maybe it'll make most sense to wait for the PMDA that uses it, and piggy-
> back on the testing of that, rather than writing a special test case just
> for the one API function.
>
> Also, please consider reclaiming the systemtap domain number for the new
> PMDA if this is is going to become the preferred stap interface for folks
> to use (and, "git rm" on the src/pmdas/systemtap - that's getting old and
> moldy at this stage. Will reports its broken on modern kernels due to it
> hooking into kernel vfs functions that haven't existed for many years, so
> it seems noone is using it anymore).
I'm not sure what the plan is (assuming there is one) for the systemtap
pmda.
--
David Smith
dsmith@xxxxxxxxxx
Red Hat
http://www.redhat.com
256.217.0141 (direct)
256.837.0057 (fax)
|