randomly get timeout error

[ikinftp]

vereion:11.0.1

I have a prgoram that run in every 30 minutes for 10 select statements. however i randomly get below error for a execution.

TimeoutError: operation timed out for an unknown reason
at /app/node_modules/tarn/dist/PendingOperation.js:17:27
at runMicrotasks ()
at runNextTicks (node:internal/process/task_queues:61:5)
at listOnTimeout (node:internal/timers:528:9)
at processTimers (node:internal/timers:502:7)

i check the statement via ssms and every query return in 1s.

[dhensby]

See #1735 (comment) - this is almost certainly not a problem with the library.

[stevendarby]

@dhensby I have read through all the related issues and some of the advice is very helpful, thank you. The issue tends to come down to badly performing queries that block connections and/or a pool size that is too small.

However, I do wonder if there is something this library could possibly do, which is to catch the tarn operation timeout errors and throw them with more contextual information and a more specific error message?

When we make a request and get this error, it's not clear if it was a timeout waiting for a connection to become available (because pool size has been reached), or creating a connection, or something else. Consumers of the mssql lib can possibly catch this error themselves and try to examine the pool status to try to determine this but I wondered if the mssql lib is able to do this more easily and provide this value to all consumers? Is that a reasonable request?

[dhensby]

That seems perfectly reasonable, there is value in knowing what part of the system has reached a timeout.

I suppose there are 3 different timeouts that can occur:

1. Timeout waiting for a connection from the pool (so this is high contention for connections from the pool, really it would indicate pool exhaustion and that consumers are not releasing connections back to the pool)
2. Timeout attempting to create a connection (the pool has capacity but actually creating the underlying connection is taking too long)
3. Timeout executing a query (the pool has available connections, the connections are open, but the query itself is what is timing out)

[stevendarby]

Thanks @dhensby. Correct me if I'm wrong, but I don't think 3 is a possibility for the error discussed here (operation timed out for an unknown reason from the tarn lib). Whenever I get 'regular' request timeouts, I get a clear SQL error about it (RequestError: Timeout: Request failed to complete in 15000ms thrown from the tedious lib). But perhaps there are certain conditions where it can throw this less clear error.

I'm still unsure about whether 2 throws this particular error as it's hard to force in a reproduction.

I've definitely been able to reproduce the error in scenario 1.

But anyway, it's precisely this uncertainty about what exact scenarios can throw this error which leads me to think catching and throwing with a bit more contextual info could be really useful. I'll raise a new item requesting that!

[dhensby]

Yes - indeed 3 is not something that would result in that specific error; I meant to enumerate all the different kinds of timeouts that could happen and the likely state that they'd happen in. To the application they are all the same thing, the request took longer to fulfil than was acceptable. But, as you say, the application (or at least logs / errors) would like a bit more context about what's actually caused the timeout and which of the 3 it is.

Thanks for the new issue :)