This message was deleted.
# atlantis-community
s
This message was deleted.
p
CPU and memory?
how big is your instance?
j
we are running i EKS, so right now its alocated 4GB memory and 4000mCPU, since the helm chart forces me to set a limit
but the real issue is not being able to run things in parallel
p
running things is parallel gets tricky
j
so this EKS cluster may have 20 applications in it
and when you go through and update them all. they are all their own project
so atlantis will go 1 by 1, doing a plan
it takes so long, the EKS token expires
so you go to apply, and it fails
p
why is takes so long? why the plans are so slow?
20 is not much
how big are your states?
j
states are tiny
they are literally just a helm chart
p
if you run one by one manually how long does it take?
j
decently quick
longest time is waiting on atlantis
p
if you tell me that plans are quick and you say 1-2 min per plan
then it should be ok, so then there is something is taking too long
I have seen CPU spikes on med size plans before
j
i have really upped the resources
p
just added a faster cpu made a huge different
j
ideally, i dont want a limit, but im stuck with one until that MR on the helm chart goes through
g
you can do
resources: null
to remove the key on the chart
although it just merged
this 1
p
I merged the PR
😎 1
j
TY!
p
have you change the 1 hour Token timeout to see if you get any further?
do you have graphs of resource usage or CPU spikes?
what do the log say?
are you using shred storage ?
j
sorry, i had stepped away
the timeout is 15 min, not 1 hour
I do have graphs, and they show throttling before i upped the specs
and yes to shared storage
p
15min?
that is very little
j
I dont think its adjustable
p
if your plans take longer than that and then it expires then you are going to have this problem for sure
no?
j
its not that the plan itself takes that long
its that if you have 20 plans
they go 1 at a time
so w/e ones were ran first, are expired, but the time atlantis updated gitlab with the entire plan
and i just copnfirmed, the EKS token is valid for 15 minute
if i have 20 helm charts, and they each take 1 minute, the first 5 expire before i can apply
p
did you always have this problem?
g
I had this problem too, and changed to requesting a token using aws-iam-authenticator. This way when the apply starts a fresh token is requested instead of using the cached in the plan.
In this example they use
aws eks token…
.
j
no chance i can get approval to use iam-authenticator
but atlantis essentialy useless at this point. Job has 10 helm changes, and its already at 12 minutes
g
j
no chance
i dont work in a normal corp env.
or in a normal, public AWS account.
g
then, with all due respect, that is not an atlantis problem
j
it is
because without atlantis, there is no problem
the fact it runs 1 by 1
g
that is a flow problem, not an atlantis problem
j
spin it how you need
g
how does it works without atlantis? you still have a 15m limit for the token
j
our other environments use flux, so they would all just sync at one time, no issues
or if i ran it locall,y i would just fire them all at the same time
same with our old env, using helmfile