That is awesome. It's amazing what a bit of optimization will do for you. On Apr 2, 2013, at 3:12 PM, fatmcgav <[email protected]> wrote:
> Yep, have been testing it today on my dev and production Puppet masters, and > seems to be working as expected... > > Have been very impressed with the performance improvements - Has dropped the > netapp_volume resource time from 40+ seconds to <0.1 seconds. So MASSIVE > improvements to be made from switching to the prefetch/flush style... > > Cheers > Gav > > > On 2 April 2013 13:24, Luke Kanies <[email protected]> wrote: > Looks good from a cursory glance. Hopefully it's all working for you. > > On Apr 2, 2013, at 9:52 AM, Gavin Williams <[email protected]> wrote: > >> As a quick update, I've pushed my latest set of code tweaks up - >> https://github.com/fatmcgav/fatmcgav-netapp/commit/7a7d18bf39cdbb04a3b0b5192929ec6a857c0a5e >> >> Need to tidy the code up a bit, but should be able to get the gist :) Have >> got a couple of defs which pull back the bulk of the information, and then a >> couple that need to be called on a volume by volume basis. >> All gets put together in instances, and provider'd up in prefetch. >> >> Comments welcome. >> >> Cheers >> Gavin >> >> On Friday, 29 March 2013 13:15:12 UTC, Gavin Williams wrote: >> Luke >> >> Cheers for that info, should prive useful... >> >> Guess I could implement a method for each property that pulls back all that >> property values for all the volumes, and then match them up in prefetch. >> >> Cheers >> Gav >> >> On 29 Mar 2013 12:49, "Luke Kanies" <[email protected]> wrote: >> Hi Gavin, >> >> Yeah, there isn't as much consistency in API as I would like. >> >> The way it should work, I believe, is something like: >> >> * The Type should support an 'instances' method that returns a complete list >> of all instances of that type that exist on the host itself. This method >> should not require a catalog. >> >> * The provider should support a 'prefetch' method, which accepts a list of >> resources from the catalog and updates them with the correct provider >> instance, with appropriate data already filled in to reflect the system >> state. >> >> * Optionally, many providers implement (I think) their own 'instances' >> method, which returns all instances and doesn't require a catalog, to >> implement both of the above methods. >> >> I just looked at the Provider base class, though, and it's not at all clear >> from this. Part of the problem is that the system is smart enough to skip >> providers that don't support prefetch, and the way we denote 'doesn't >> support prefetch' is that it doesn't have the method. Given that, I left >> the method off of the base class. Because 'instances' isn't used for this >> kind of test, the base class ships with a stub method that just throws an >> error. >> >> So yeah, all that's confusing, and could really use some additional work to >> make it easier to understand, and to bring some consistency. >> >> On Mar 29, 2013, at 12:37 PM, fatmcgav <[email protected]> wrote: >> >>> Luke >>> >>> No worries. >>> >>> For the 2 I've converted soo far, it's made quite a performance difference. >>> Thankfully can pull back all the resources with one NetApp api call, so >>> it's drop the resource time from 5+ seconds to sub 1 second. >>> However was a bit confused initially, as different providers seem to use >>> different models... Some with just prefetch, some with just instances, some >>> with both... >>> >>> However I'm just about to start the more complex netapp_volume provider, >>> which has 4+ properties that have different api calls on top of the base >>> resource list call. This is by far the slowest provider out of them with an >>> average of 20+ seconds per run. >>> So not sure of the best way to handle as yet. >>> >>> Cheers >>> Gav >>> >>> On 29 Mar 2013 12:11, "Luke Kanies" <[email protected]> wrote: >>> Hi Gavin, >>> >>> I'm glad to see you've sorted it out, sorry I didn't jump in before it was >>> all resolved. >>> >>> How has prefetching worked out for you, in terms of performance? >>> >>> On Mar 28, 2013, at 3:28 PM, Gavin Williams <[email protected]> wrote: >>> >>>> Afternoon >>>> >>>> Managed to find my issue.. Variable name re-use :( >>>> >>>> Defined 'qtrees' as an empty array on line 3, and then populated it with a >>>> whole load of device output on line 18 :( >>>> >>>> So have defined a 'qtree_instances' array on the outside to contain my >>>> output, with qtrees being used on the inside to hold the NetApp filer >>>> response... Latest code in Github. >>>> >>>> Cheers >>>> Gavin >>>> >>>> On Thursday, 28 March 2013 11:44:39 UTC, Gavin Williams wrote: >>>> Morning all >>>> >>>> Quick update... Looks like I managed to hack around the issue by adding >>>> the following: >>>> >>>> ... >>>> ap qtree_info >>>> >>>> # Check if it is a NaElement >>>> >>>> >>>> >>>> next unless qtree_info.respond_to?(:child_get_string) >>>> >>>> # Pull out the qtree name. >>>> >>>> >>>> >>>> name = qtree_info.child_get_string("qtree") >>>> ... >>>> >>>> However this shows that I'm getting a total of 78 items processed, whereas >>>> the original array only contains 53 items... >>>> The additional items being processed are all like: >>>> '#<Puppet::Type::Netapp_qtree::ProviderNetapp_qtree:'. >>>> Have updated the gist with latest code and log file. >>>> >>>> Would like to understand where these are coming from, and if it's >>>> something I'm doing incorrectly? >>>> >>>> In the mean-time, following fixing that bug, the provider now seems to >>>> work as expected :) >>>> >>>> Cheers >>>> Gavin >>>> >>>> >>>> On Wednesday, 27 March 2013 17:32:23 UTC, Gavin Williams wrote: >>>> Afternoon all >>>> >>>> I've started working on converting a couple more of my NetApp network >>>> device providers to use a prefetch/flush model, as can see performance >>>> gains available, etc... >>>> >>>> Anyways, I'm having issues with my netapp_qtree provider. It would appear >>>> that somehow, an additional Puppet::Type... row is getting into an array >>>> and breaking things... >>>> >>>> Have created a gist here with the details, as the log file is quite long. >>>> Also includes the instances and prefetch def's for my netapp_qtree >>>> provider... >>>> >>>> As you can see on Line 337 of the log, the array contains 40 items, >>>> however on line 734 self.instances is trying to process item 41?!?! >>>> What's also strange is that the item contents look like a Puppet Type >>>> (#<Puppet::Type::Netapp_qtree), whereas all the others in the array are >>>> NetApp specific items (#<NaElement:). >>>> >>>> So, any ideas??? >>>> >>>> Cheers >>>> Gavin >>>> >>>> -- >>>> You received this message because you are subscribed to the Google Groups >>>> "Puppet Developers" group. >>>> To unsubscribe from this group and stop receiving emails from it, send an >>>> email to [email protected]. >>>> To post to this group, send email to [email protected]. >>>> Visit this group at http://groups.google.com/group/puppet-dev?hl=en. >>>> For more options, visit https://groups.google.com/groups/opt_out. >>>> >>>> >>> >>> >>> -- >>> Luke Kanies | http://about.me/lak | http://puppetlabs.com/ | +1-615-594-8199 >>> Join us at PuppetConf 2013, August 22-23 in San Francisco - >>> http://bit.ly/pupconf13 >>> >>> >>> -- >>> You received this message because you are subscribed to a topic in the >>> Google Groups "Puppet Developers" group. >>> To unsubscribe from this topic, visit >>> https://groups.google.com/d/topic/puppet-dev/74JW491YSAk/unsubscribe?hl=en. >>> To unsubscribe from this group and all its topics, send an email to >>> [email protected]. >>> To post to this group, send email to [email protected]. >>> Visit this group at http://groups.google.com/group/puppet-dev?hl=en. >>> For more options, visit https://groups.google.com/groups/opt_out. >>> >>> >>> >>> -- >>> You received this message because you are subscribed to the Google Groups >>> "Puppet Developers" group. >>> To unsubscribe from this group and stop receiving emails from it, send an >>> email to [email protected]. >>> To post to this group, send email to [email protected]. >>> Visit this group at http://groups.google.com/group/puppet-dev?hl=en. >>> For more options, visit https://groups.google.com/groups/opt_out. >>> >>> >> >> >> -- >> Luke Kanies | http://about.me/lak | http://puppetlabs.com/ | +1-615-594-8199 >> Join us at PuppetConf 2013, August 22-23 in San Francisco - >> http://bit.ly/pupconf13 >> >> >> -- >> You received this message because you are subscribed to a topic in the >> Google Groups "Puppet Developers" group. >> To unsubscribe from this topic, visit >> https://groups.google.com/d/topic/puppet-dev/74JW491YSAk/unsubscribe?hl=en. >> To unsubscribe from this group and all its topics, send an email to >> [email protected]. >> To post to this group, send email to [email protected]. >> Visit this group at http://groups.google.com/group/puppet-dev?hl=en. >> For more options, visit https://groups.google.com/groups/opt_out. >> >> >> >> -- >> You received this message because you are subscribed to the Google Groups >> "Puppet Developers" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected]. >> To post to this group, send email to [email protected]. >> Visit this group at http://groups.google.com/group/puppet-dev?hl=en. >> For more options, visit https://groups.google.com/groups/opt_out. >> >> > > > -- > Luke Kanies | http://about.me/lak | http://puppetlabs.com/ | +1-615-594-8199 > Join us at PuppetConf 2013, August 22-23 in San Francisco - > http://bit.ly/pupconf13 > > > -- > You received this message because you are subscribed to a topic in the Google > Groups "Puppet Developers" group. > To unsubscribe from this topic, visit > https://groups.google.com/d/topic/puppet-dev/74JW491YSAk/unsubscribe?hl=en. > To unsubscribe from this group and all its topics, send an email to > [email protected]. > To post to this group, send email to [email protected]. > Visit this group at http://groups.google.com/group/puppet-dev?hl=en. > For more options, visit https://groups.google.com/groups/opt_out. > > > > > -- > You received this message because you are subscribed to the Google Groups > "Puppet Developers" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To post to this group, send email to [email protected]. > Visit this group at http://groups.google.com/group/puppet-dev?hl=en. > For more options, visit https://groups.google.com/groups/opt_out. > > -- Luke Kanies | http://about.me/lak | http://puppetlabs.com/ | +1-615-594-8199 Join us at PuppetConf 2013, August 22-23 in San Francisco - http://bit.ly/pupconf13 -- You received this message because you are subscribed to the Google Groups "Puppet Developers" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/puppet-dev?hl=en. For more options, visit https://groups.google.com/groups/opt_out.
