a year ago
Hi
I did a test where I deployed a rust axum rest api twice, one on Metal and one on GCP. But the latency on Metal is just way higher.
The increased latency seems to be coming from the database connection.
I apologise in advance if I did something stupid :)
36 Replies
a year ago
did you also put the database on metal?
a year ago
is your metal service using the metal edge as well?
I have tried switching between the metal edge and the old one but doesnt seem to make that much of a difference
a year ago
its a DNS update so please be patient, how long has the domain been on the metal edge?
Well ive been switching between them pretty quickly. Just switched back to the Metal edge. How long do you recommend waiting before testing again?
a year ago
you can dig the domain, when it starts returning a list of IP addresses you should be good
a year ago
dig, the cli tool
Is it this you're looking for?
edge.railway.app. 152 IN A 66.33.22.3
edge.railway.app. 152 IN A 66.33.22.1
edge.railway.app. 152 IN A 66.33.22.4
edge.railway.app. 152 IN A 66.33.22.2a year ago
yep thats the metal edge
a year ago
can you run a traceroute to one of the ip addressees?
traceroute to 66.33.22.1 (66.33.22.1), 30 hops max, 60 byte packets
1 172.20.64.1 (172.20.64.1) 0.222 ms 0.151 ms 0.164 ms
2 mymodem (192.168.1.1) 0.707 ms 1.157 ms 0.891 ms
3 10.24.49.76 (10.24.49.76) 9.628 ms 9.678 ms 10.193 ms
4 * * *
5 ae-64-100.ibrstr6.isp.proximus.be (91.183.241.200) 17.692 ms 17.436 ms 19.219 ms
6 brx-b4-link.ip.twelve99.net (213.248.97.112) 19.279 ms 15.113 ms 15.188 ms
7 adm-bb2-link.ip.twelve99.net (62.115.116.230) 17.139 ms 15.470 ms 15.726 ms
8 adm-b12-link.ip.twelve99.net (62.115.137.191) 16.426 ms 17.138 ms 17.142 ms
9 railwaycorp-ic-390073.ip.twelve99-cust.net (62.115.196.223) 18.081 ms 18.836 ms 17.939 ms
10 * * *
11 * * *
12 * * *
13 * * *
14 * * *
15 * * *
16 * * *
17 * * *
18 * * *
19 * * *
20 * * *
21 * * *
22 * * *
23 * * *
24 * * *
25 * * *
26 * * *
27 * * *
28 * * *
29 * * *
30 * * *a year ago
and if you could please visit https://rlwy-edge-info-production.up.railway.app/ and paste the response body here + your current location
edge=railway/europe-west4-drams3a
zone=europe-west4-drams3a
ip=::ffff:100.64.0.2
forwarded=91.177.28.24
hs=f09480be9e4a
req=T6a7ncZgTVaGcj5738lYAA_3554946428How accurate should my current location be? :)
a year ago
as accurate as you feel comfortable announcing publicly
a year ago
fr?
a year ago
you know these threads are indexed publically on google?
a year ago
fair enough
a year ago
cc @Phineas - slightly slower response times on Metal.
✅ - db on metal
✅ - app on metal
✅ - region same as user
✅ - metal edge enabled
✅ - priv net for db comms
I can't think of anything else that could cause issues.
Side note:
Even when querying an endpoint with no database interaction its still slightly slower.
GCP:
a year ago
interesting, ill let phin weigh in on this, so lets wait to hear back from him
Not sure if you guys looked any further into this but I just tested again and this seems to mostly be fixed. GCP still seems to be faster but its by a difference of <5ms on average.
9 months ago
so sorry that this thread got lost, we made some changes and are seeing metal response times faster from our testing data now