goerz / gc3pie

Automatically exported from code.google.com/p/gc3pie
0 stars 0 forks source link

Detect when reaching the limit of available VMs #397

Open GoogleCodeExporter opened 9 years ago

GoogleCodeExporter commented 9 years ago
We need to have a mechanism to detect when new VMs cannot be launched 
any-longer.
The current behaviour is to keep trying indefinitely
while, probably, it should be possible to detect a failure due to exhausted 
quota.

Also when a system failure occurs (like no more public IPs)
it should be possible to detect it and avoid trying again (at least for the 
current 'progress' cycle)

Sergio :) 

Original issue reported on code.google.com by sergio.m...@gmail.com on 18 Mar 2013 at 10:06

GoogleCodeExporter commented 9 years ago
AFAIK, there is no API to get the maximum number of instances from AWS, so boto 
has no support for it, so we cannot know if we reached our limit.

Moreover, even if you were able to know that the creation of an instance failed 
because there were no more free IPs available (which is something that should 
never happen, unless you made some mistakes designing your network) (or your 
network administrator does not give you enough IP addresses...), you couldn't 
know when the problem will be solved.

Therefore, since we cannot possibly know if the error we get while creating the 
instance is temporary or permanent, and we couldn't do anything to fix it 
anyway, the current behaviour (delegate to the Engine class and try to submit 
everytime you can) is the correct one. The only thing we may need to check is 
if we are correctly terminating the instance in error state and, if we aren't 
terminating it, if this can cause any waste of resources (I don't know if an 
instance in error state counts as 1 running instance while computing the 
current used quota for the tenant)

P.S.
If we will ever implement an OpenStack backend, then yes, we will be able to 
get the current quota and update the available resources accordingly.

Original comment by arcimbo...@gmail.com on 18 Mar 2013 at 10:37

GoogleCodeExporter commented 9 years ago
| AFAIK, there is no API to get the maximum number of instances from
| AWS, so boto has no support for it, so we cannot know if we reached
| our limit.

But we can scan the error message from Boto and extract the information from it?

Original comment by riccardo.murri@gmail.com on 19 Aug 2014 at 8:54