-
Notifications
You must be signed in to change notification settings - Fork 1.1k
Make the chaos type pluggable #86
Conversation
Shutting down an instance is just one way an instance can fail. This patch makes it pluggable, so that we can simulate more failure types.
|
SimianArmy-pull-requests #44 FAILURE |
|
SimianArmy-pull-requests #47 FAILURE |
|
SimianArmy-pull-requests #48 FAILURE |
|
SimianArmy-pull-requests #49 FAILURE |
We had a circular initialization loop (at least in theory)
Whoops ... pains of doing a manual merge!
|
SimianArmy-pull-requests #51 SUCCESS |
|
SimianArmy-pull-requests #52 SUCCESS |
|
SimianArmy-pull-requests #53 FAILURE |
|
SimianArmy-pull-requests #54 FAILURE |
|
tests are failing: |
|
SimianArmy-pull-requests #55 SUCCESS |
Now that we can enable/disable chaos types, the field name was misleading.
|
SimianArmy-pull-requests #56 SUCCESS |
|
SimianArmy-pull-requests #57 SUCCESS |
|
Sorry for failures... while developing the actual plugins I was adding functionality into the base classes. I've got two new plugins now (one active pull request #87 , one that needs a bit more testing), so I think this "refactor for pluggability" PR should be OK now. |
|
Hi Justin, Thanks for the contributions. The changes look interesting. Sorry it has taken so long for me to look at this. The pull requests are sort of a mess with the cumulative changes mixed in and basically too hard for me to find the meaningful content. I think it would make more sense to merge all the various chaos type changes into one large pull request. Also I don't think I saw any unit tests provided? Perhaps I just missed them, but getting some tests added would really be appreciated. Thanks |
|
No worries Cory! The Pull requests are more or less cumulative, as most of them build on the previous ones. I definitely agree they don't make any sense if read out of order! I think the order should be the numerical order in Github. If I look at this PR #86 in this view: https://github.com/Netflix/SimianArmy/pull/86/files It seems fairly straightforward (though I'm obviously biased!); this PR is setting up the infrastructure for allowing the chaos type to be pluggable, pulling the shutdown instance strategy into its own class. There are a lot of commits that github is collapsing though; let me know if you'd like me to rebase (or do some other git magic!). I'll look at adding unit tests. Do you have any suggestions for what would be a good unit tests to add here and on the others PRs? I've mostly been relying on the existing test suite along with manual testing (it's tricky to do a lot of this without real EC2 instances)... |
|
Hey, it looks like pull request #93 is comprehensive of all the requests right? I think I will just review that more with more detail and add comments if I seen anything I think needs changing. For the unit tests I was going to point you to our coverage reports, but the reports seem broken at the moment: I will try to get the coverage reports working again then that should highlight any chunks that need additional testing. Thanks, |
|
Coverage reports would be great to work from. #93 should be everything, so if you want to see the final picture then by all means jump in there! I tried to keep them as individual steps in case you wanted to go in small chunks, but (as you can see) I think it made the Git history a lot messier. If you go into my repo I think it's easier to compare the individual branches that back each pull request, if you want to do that. Whatever review style works for you, works for me :-) |
Shutting down an instance is just one way an instance can fail.
This patch makes it pluggable, so that we can simulate more failure types.