AutoScale formula doesn't work - $PendingTasks.GetSample(...) - returns still "0"

Tomas Marsak 6 Reputation points
2022-02-01T13:00:26.877+00:00

Hello,

we have problem with AutoScale formula, it looks that it stop working during yesterday evening. We have new Jobs (with tasks) registered / visible on Azure Batch account, but AutoScale formula still returns 0 PendingTasks.

AzureBatch Account: nnpolarproduction
Pools: Pool_d16_v3_5 / Pool_d8_v3_40

If I will try to "evaluate" only following line in formula:
$pendingTasks=$PendingTasks.GetSample(10000);
Result is:
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]

But we have Jobs (tasks) scheduled / visible on the "Jobs" tab in azure portal, Jobs are scheduled on correct pools.

Any know issue with Azure Batch & AutoScale formula ?? Or any idea what we should check ??

Thank you

Azure Batch
Azure Batch
An Azure service that provides cloud-scale job scheduling and compute management.
320 questions

2 answers

Sort by: Most helpful
Most helpful Newest Oldest
  1. Tomas Marsak 6 Reputation points
    2022-02-02T16:43:32.823+00:00

    Confirmed by MS Support it is incident in WestEurope region. Formulas doesn't work.

    Suggested workaround:

    • use NorthEurope region, by the information from MS Support this region shoudln't be affected

    There is no ETA for that issue in WestEurope :-\

    This is pretty stupid, you have to either manually manage the resources in the Pool, or keep all the nodes running permanently and then pay the extreme bill.

    I'm quite surprised that MS has no SLA.

    0 comments
  2. i_am_xcfje 0 Reputation points
    2023-11-06T14:27:03.7266667+00:00

    We have a pool in Western Europe which seems to have the same behavior.

    If we submit a job with 20k tasks then $PendingTasks is not updated until all the tasks are submitted (even though Azure Batch has no knowledge of how many tasks there will be eventually).

    If we submit a second parallel job with 200 tasks then $PendingTasks only reflects this job's tasks (possibly because it was submitted very quickly in comparison, and reaches that "all tasks submitted" state faster).

    @vipullag-MSFT please confirm if this issue in still open nearly 2 years later?