could you get me your account id please

could you get me your account id please? so that I can look into this futher

7 Replies

hey Luis! Thanks for the response – I wouldn't waste too much time on it, I'm planning on adding lots of logs tonight and seeing what I can find. 528d1f395656d35eea56f99692160727 – you can see a good example on ID meas_p6r5. I've removed the parallel a few days ago already, the workflow itself is something like the image with 3 steps. Yesterday I added some logs and re-deployed and my first impression was that the someFunction was working but something was rather getting disconnected on the worker=>workflow=>call. Or that it'd be really weird for 120s not to be a long enough window persistently in 5 retries (and the upstream services seemed all OK). Today did a few runs and they worked on the first attempt in 15-25s

SeekerdasbatatasOP•2w ago

and just to confirm the way you were making it parallel was by using a Promise.all?

juanferreras•2w ago

that was unrelated to these runs though (e.g. it just caused a similar defect where I was seeing them timeout and had no clue what was going on, and I have the feeling that the upstream APIs were returning OK but something getting disconnected along the way or not awaiting). I wonder if my case of not really returning intermediate data (the async function just writes partial results in the DB already) might be giving me a different experience than intended

if (runFirst) {
  const firstPromise = (async () => {
    try {
      await step.do("fetch-first-data", STEP_CONFIG, async () => {
         ...
    } catch
        ...
  })();
  stepPromises.push(firstPromise);
}

await Promise.all(stepPromises);

if (runFirst) {
  const firstPromise = (async () => {
    try {
      await step.do("fetch-first-data", STEP_CONFIG, async () => {
         ...
    } catch
        ...
  })();
  stepPromises.push(firstPromise);
}

await Promise.all(stepPromises);

hmm seeing the code again makes me think there might be better ways to do that too

SeekerdasbatatasOP•2w ago

hum I see, I don't see anything inherently wrong with doing it like this - do you remember what was the timeout in these steps? the same 2 minutes as the picture above? I'll try to investigate this later, but the error you received is supposed to be transient? Did you see it consistently in all attempts?

juanferreras•2w ago

it was 10 mins haha until I realized it was per-attempt and not total 😂

juanferreras•2w ago

tried the parallelism again and it seems to work reliably haha – I think I'll make sure to add a lot of logs next time I see this issue to finally hunt it down I think this can be discarded now – I've done a bunch of changes, and mainly split workflow steps in smaller units of work (with less subrequests/parallel subrequests) and it's working a lot more reliably now. thanks Luis!

SeekerdasbatatasOP•2w ago

lmk if it happens again :MeowHeartCloudflare:

Gaming

Programming

could you get me your account id please

Did you find this page helpful?