2 months ago
Are there any performance benefits selecting a postgres database vs deploying a postgres from a docker image on railway?
I have deployed the gel database template on railway and seeing degraded performance compared to the geldata cloud (which is hosted by gel).
Wondering if I should rather select a postgres database on railway and host a gel server next to it, if that would boost performance.
11 Replies
2 months ago
Hey there! We've found the following might help you get unblocked faster:
If you find the answer from one of these, please let us know by solving the thread!
2 months ago
It really depends on your use case. For most apps, using Postgres in Railway will be good enough in terms of performance.
However, there are other implications running on Railway (self-managed) vs any other managed database and it's that you must take care of it. If you want to backup or upgrade versions, you'll need to do it yourself. There are a couple templates to make this less complicated like https://github.com/railwayapp-templates/postgres-s3-backups but still doing a recovery would be manual work.
If you feel comfortable doing that, then I suspect hosting on Railway would be more cost-effective and performant enough.
If you host the Gel server next to the Postgres, it will be more performant than running the server at a different place (esp. if not same region) so you can try that out.
crisog
It really depends on your use case. For most apps, using Postgres in Railway will be good enough in terms of performance.However, there are other implications running on Railway (self-managed) vs any other managed database and it's that you must take care of it. If you want to backup or upgrade versions, you'll need to do it yourself. There are a couple templates to make this less complicated like https://github.com/railwayapp-templates/postgres-s3-backups but still doing a recovery would be manual work.If you feel comfortable doing that, then I suspect hosting on Railway would be more cost-effective and performant enough.If you host the Gel server next to the Postgres, it will be more performant than running the server at a different place (esp. if not same region) so you can try that out.
2 months ago
Okay. Im fine with upgrading myself.
I am just a bit surprised to see the startup times of queries 4x slower on railway compared to the gel cloud. Even though on railway i have way more resources. Configs are exactly the same on both instances
2 months ago
Railway doesn’t offer managed Postgres, the database thing is just a “template loader”
2 months ago
Also make sure the db and gel are in the same instance
2 months ago
Yes I think they are - I have used the gel railway template which uses gel docker image
2 months ago
Check if they are
nikolandgraf
Yes I think they are - I have used the gel railway template which uses gel docker image
2 months ago
I'll try this myself and report back
nikolandgraf
Yes I think they are - I have used the gel railway template which uses gel docker image
2 months ago
I managed to deploy this with an standalone PostgreSQL server + Gel container. Please try it and let me know if you see any performance improvements.
https://railway.com/deploy/gel-postgresql
You can also tweak GEL_SERVER_MAX_BACKEND_CONNECTIONS and GEL_SERVER_COMPILER_POOL_MODE to improve performance. Related docs,
https://docs.geldata.com/reference/running/configuration#gel-server-compiler-pool-mode
https://docs.geldata.com/reference/running/configuration#gel-server-max-backend-connections
2 months ago
Alright - deploying the postgres as a separate service next to the gel server made all the difference!
Now performance looks roughly equal.
So would not recommend to use this template here: https://railway.com/deploy/3tpcCB
nikolandgraf
Alright - deploying the postgres as a separate service next to the gel server made all the difference! Now performance looks roughly equal. So would not recommend to use this template here: https://railway.com/deploy/3tpcCB
2 months ago
Glad to hear it's working better now!
If you ever need more help don't hesitate to ask 🫡