Hey guys! I'm facing a problem with keeping SuperTokens's process up. I have a remote EC2 instance on AWS, I run "sudo supertokens start", and it works for a while. After some time, the process stops. I tried to create a service:

[Unit]
Description=Supertokens
After=network.target

[Service]
Type=simple
User=root
ExecStart=/usr/bin/bash -c "sudo /usr/bin/supertokens start"
ExecStop=/usr/bin/bash -c "sudo /usr/bin/supertokens stop"

[Install]
WantedBy=multi-user.target

But the situation is worse: the service stops immediately, and if i put the

Restart: 'always'

option, it basically restarts SuperTokens endlessly *(consuming tons of CPU) *. I'm not using Docker!

Hey!

Yes, with --foreground the service does actually start and work. I don't know how long does it lasts (it usually stops working the next day), so I'll let you know tomorrow if something happens. (I've used this guide https://gist.github.com/CuriousCI/8106c780a4a95f835cba2ead599839df to create the service)

hmm. There have been no such issues that we have seen when running the core ourselves. And we have been running quite a few without restarting it for > 1 year even,

It might also be a problem with the EC2 AWS instance I'm using, which is very limited in memory and CPU

hmm. I have run the core using t2.micro as well - for a very long time

It might be the fact that I'm running the DB and the python backend too... It happened again right now, and I'm checking the errors

hmm okay.

Ok, the Java process is the problem... the PID of the process killed is the same as the one SuperTokens got, so yes... that might be one of the problems... the other one is yum, but it's not related to SuperTokens...

$ dmesg
...
[12835829.721863] Out of memory: Kill process 11750 (java) score 210 or sacrifice child
[12835829.728337] Killed process 11750 (java) total-vm:2122848kB, anon-rss:105756kB, file-rss:0kB, shmem-rss:0kB
...
[12835832.563089] Out of memory: Kill process 4917 (yum) score 439 or sacrifice child
[12835832.568976] Killed process 4917 (yum) total-vm:536324kB, anon-rss:221104kB, file-rss:0kB, shmem-rss:0kB

$ cd /usr/lib/supertokens/.start
$ cat 0.0.0.0-3567

11750

Hmm i see. 50 mb might be a little too less

The logs give some information... something happened this morning at 3 AM... let me get you the files

it takes a min of 180 MB and on normal load, can go upto 400 MB of RAM. So i think you might wanna give it 300 MB just to be safe

Something strange happens at 3 AM today according to the logs

error.log is empty?

I think the problems isn't the process not having enough memory, but the system killing the process because it consumes too much memory

hmm

I'll try adding some swap memory

let me check as well about the core's behavior with 50 mb limit

The strange thing is that the process behaves normally for 2 days, then it decides to explode

that is strange indeed. Is there a spike in requests? Which version of the core are you using?

The metrics don't show any spike in requests, I'm using core 3.12.1

hmm. that should not have any memory leak issue. Are you using argon2 hashing by any chance?

Yeah, I use 282 MB by default without SuperTokens running, it is a problem of machine's memory

so no argon2?

No argon2

I see.

I'll check if there are any problems with swap memory enabled, and I'll let you know

ok thanks! i'll test on our side as well.

Ah, another thing: when the system stops the process, I have to manually delete the file in the

.start

folder, or the SuperTokens CLI will give some problems.

ah yea. that;s expected

Amazon Linux 2

Ok great!