I'm getting connect ETIMEDOUT errors when Im tryin...
# orm-helpr
rein
05/02/2018, 12:41 PMI'm getting connect ETIMEDOUT errors when Im trying to pass a larger number of mutations(500) to my zeit now endpoint. Could someone perhaps tell me what I am doing wrong? This used to be no problem at all when using Graphcool Framework
h
huv1k
05/02/2018, 12:42 PMSmaller batches?
r
rein
05/02/2018, 12:46 PMIsn't there some automatic batching? Its 500 create mutations
m
Moritz
05/02/2018, 1:07 PMDo these mutations access the prisma backend? The free prisma service is limited to a certain bandwidth (I think about 1 request/s). They cache some requests and are somewhat flexible resulting in small delays, but at some point the cache limit is reached and the prisma service stops responding. I solved the same issue by just adding a 1 second timeout in my scripts after each mutation. Takes some time but does the job 🙂 If it bothers you consider hosting your own prisma instance in docker.
r
rein
05/02/2018, 1:08 PMIm paying for prisma cloud, so I dont think it is rate limited
rein
05/02/2018, 1:08 PMor is it also for paying prisma cloud users?
h
huv1k
05/02/2018, 1:10 PMYea, there is limit
m
Moritz
05/02/2018, 1:10 PMIm sure they have some kind of rate limits to manage their infrastructure, but they should be reasonably high for paying user. 500 mutations should not be a problem for a paying user 🙂 Dont have any numbers though
r
rein
05/02/2018, 1:10 PMthats what I thought
h
huv1k
05/02/2018, 1:11 PMn
nilan
05/02/2018, 1:11 PMIs
ETIMEOUTnowr
rein
05/02/2018, 1:11 PMit is
n
nilan
05/02/2018, 1:12 PMwhat happens if you send 500 other HTTP requests?
nilan
05/02/2018, 1:12 PMsounds like
nowm
Moritz
05/02/2018, 1:14 PMFor me on the free plan, the same scenario occured when bombing the prisma endpoint directly 🙂
n
nilan
05/02/2018, 1:14 PMthrough
nowr
rein
05/02/2018, 1:14 PMBut I am not on the free plan though, I have my own cluster
n
nilan
05/02/2018, 1:14 PMI never saw a Prisma API return
ETIMEDOUTr
rein
05/02/2018, 1:15 PMyeah its the request itself that fails
rein
05/02/2018, 1:15 PMnot the prisma endpoint
rein
05/02/2018, 1:15 PMI also never had any troubles while doing this with graphcool framework
m
Moritz
05/02/2018, 1:16 PMLet me reproduce again to double check 🤒 I have a local seeding script written in JS directly talking to my prisma endpoint.
n
nilan
05/02/2018, 1:16 PMThat's quite irrelevant 🙂 You also never used
nownilan
05/02/2018, 1:17 PMWould be great if you can learn more about this @rein, for example try to use a delay between mutations as @Moritz suggested
r
rein
05/02/2018, 1:18 PMIll let you know!
n
nilan
05/02/2018, 1:19 PMYou can also look into your DB metrics when doing those queries/mutations
nilan
05/02/2018, 1:22 PMif you see something suspicious there, you can try the same with a beefier DB
r
rein
05/02/2018, 1:23 PMI dont think its that big of a load, but who knows
rein
05/02/2018, 1:23 PMIt worked fine on graphcool framework on a free plan
rein
05/02/2018, 1:23 PMso
rein
05/02/2018, 1:23 PMbut ill do some testing
h
huv1k
05/02/2018, 1:23 PMbecause it went trought different infrastructure
n
nilan
05/02/2018, 1:23 PMlet's gather data, then we know more 🙂
h
huv1k
05/02/2018, 1:24 PMzeit made some changes to their infrastructure because a lot of DDOS
huv1k
05/02/2018, 1:24 PMtry to host your graphql server on different provider maybe it will handle okey 🙂
m
Moritz
05/02/2018, 1:38 PM@nilan Haha ok so I reran my script and everything went through fine this time, even without any delays between mutations. Prisma seems to have been doing some fitness 🙂 2800 requests in 15 minutes on the free plan without any issues. My issue before might thus have been of a local or connection nature like you suggested. Nice performance, props! 🙌
👌 1
Moritz
05/02/2018, 1:47 PM@rein You can directly call mutations on your prisma endpoint via a script as such. (If you want to avoid going through
now2 Views