9 months ago
performace on metai is much worse for me, most likely biggest part is PSQL side, i dont have decent monitoring to know for sure. One of background tasks that took 15sec (scr1) right before transition to metal, right after it takes 75-80sec (scr2). This task relies on PSQL a lot. Ofc i didnt deploy any changes in code between these 2 screenshots. Because of this terrible performance some endpoints responses take 1-2sec, which didnt happen before transition
ive tried migrating back to EU legacy on dev, same code, task completion time came back to 15sec. Ive waited for sunday to do the same on production, but now this option doesnt show. What do i do?
also transition to metal was done in production without my consent or even warning. We have an important system which relies on static IP, IP changed, hence system didnt work for few hours. Even better - after PSQL migration backend server wasnt restarted for 3 hours, and didnt work at all. Why woulld you do such a thing, whats the idea? even if you force thansition, why not give me a chance to do it controlled, by myself?
UPD: great, dev environment got auto-migrated to metal too, and now simple endpoints with like 20 objs from PSQL take 3-20seconds to load. Not sometimes - always. Will be even better when production reaches the same level
11 Replies
9 months ago
screenshots ive mentioned. right before and right after metal
Attachments
8 months ago
great, dev environment got auto-migrated to metal too, and now simple endpoints with like 20 objs from PSQL take 3-20seconds to load. Not sometimes - always. Will be even better when production reaches the same level
Attachments
8 months ago
Hello!
We've escalated your issue to our engineering team.
We aim to provide an update within 1 business day.
Please reply to this thread if you have any questions!
Status changed to Awaiting User Response Railway • 9 months ago
8 months ago
Hello,
The extra latency you are perceiving here is caused by accessing Postgres and Redis over the public network, that path involves more hops and thus extra latency.
To resolve this, please use the private network exclusively for all database to service communications.
Best,
Brody
8 months ago
Yes, switch to private networking helped, thanks. Also please make docs more clear on how to do this - searching that correct port of service for private networking is called RAILWAY_TCP_APPLICATION_PORT is really not easy for people who dont navigate well in these network protocols and such. Maybe ive missed something but i dont see it explicitly said in docs
Status changed to Awaiting Railway Response Railway • 9 months ago
Status changed to Solved strellic9 • 9 months ago
8 months ago
🛠️ The ticket Performance issue with disk operations on metal has been marked as todo.
5 months ago
✅ The ticket Performance issue with disk operations on metal has been marked as completed.
5 months ago
🛠️ The ticket Performance issue with disk operations on metal has been marked as in progress.
5 months ago
✅ The ticket Performance issue with disk operations on metal has been marked as completed.
5 months ago
🛠️ The ticket Performance issue with disk operations on metal has been marked as in progress.
5 months ago
🛠️ The ticket Performance issue with disk operations on metal has been marked as in progress.
5 months ago
✅ The ticket Performance issue with disk operations on metal has been marked as completed.