Hi folks, is anyone seeing unusually
Hi folks, is anyone seeing unusually high latency right now? I am seeing issues in prod when making DB queries. I'm on the latest wrangler versions, and I haven't had issues w Hyperdrive recently
12 Replies
Hey are you seeing any errors related to these higher latencies?
i'm getting a 502 gateway error after hanging for 2min
i don't think it's the underlying database that's the issue
There are a couple of 502 bad gateway errors on our end but not seeing any global latency increases. I'll keep looking around to see if theres anything unusual. We're in the middle of a revert of a previous release
do you happen to have a hyperdrive config ID?
16984f8fed7f4d428135944bd92240fc
for remote dev
this is what our hyperdrive dashboard looks like
Hey @kchro3 we fully released our revert late last night. Are you still seeing any increased latencies? I noticed that your configuration has seen some improvements, just want to verify from your end everything is back to normal
yes, it does seem to load faster now. would you be able to share more about what was the root cause?
would be curious to know if there are limits on the size of payloads etc that we should be aware of because it was slower for larger payloads
another general question is that we are seeing lots of "uncacheable" requests, but we are mostly sending read-only queries.
is it because we're using template variables & it can't cache those?
We're still determining the root cause but we're in the middle of rolling out a critical change to the hyperdrive architecture that will allow for better caching closer to the user vs our current architecture which is caching closer to the database. The final result of this large effort should greatly improve cached latency results to users
We do cache parameterized queries I would have to think a bit further would could be the cause of the missed caches
We don't cache anything within a transaction.
Anything with volatile functions in it (things like random or sleep or such) also can't be safely cached.
My guess is it's one of those two.
My guess is it's one of those two.
the thing that was super slow yesterday was when we were sending a batch of a complex select statements with multiple joins & params. the params themselves are simple strings (low cardinality)
we could probably improve performance on our side by fetching all the data in one query rather than batching, but it hadn't been a problem until yesterday
There was some issue with yesterday's release that we still need to RCA. If you're willing to DM me your queries that would likely help us as we work out why some traffic (including yours) was impacted when most was not.
yep, will dm