Container exited unexpectedly with code 0x0000DEAD

[Anonymous]

Just checking to see whether other users of Premium are experiencing this error intermittantly over the last couple of days?  I've had a few refreshes fail with this message.  Subsequent refreshes of the models work fine, so its not a big deal (so far).

Our capacity is in South-East Asia, in case thats relevant to other users also having problems.  So far there is nothing relevant listed on the support page here:

https://powerbi.microsoft.com/en-us/support/

[rubymaya]

Hello, 

I had this same error with my dataflow refresh today :

PipelineException: Container exited unexpectedly with code 0x0000DEAD. PID: 6692.

The dataflow is in the same workspace as the reports. Is there any interuptions?

[V-lianl-msft]

Hi @Anonymous ,

Does the issue happen when you refresh the dataflow? 

Please try below suggestions:

There is a semantic difference when the entities are referenced from DF in same workspace vs DF is a different workspace with respect to the compute engine. In the same workspace case, the dataflows have a strong reference to each other and are updated in the same transaction. Hence we do not need to cache data and can refer to the data from the upstream entity. However, when they come from different workspace, the references are weak references and in order to be self-contained within a workspace we do need to re-cache the data. The re-caching step is what adds the additional time in processing.

To mitigate this we suggest

1. Increase dataflows’ max memory % and decrease the Datasets max memory % by same quantity. (If they have default settings then increase dataflows to 40% and decrease Datasets to 80%)
2. Increase Dataflows container size to 1500Mb. This will reduce parallelism and potentially reduce intermittent failures.

Best Regards,
Liang
If this post helps, then please consider Accept it as the solution to help the other members find it more quickly.

[Anonymous]

Hi @V-lianl-msft   we don't use dataflows as we haven't found a use case that made them superior to power query, for our needs.  The issue hasn't persisted in the last 24 hours, so whatever was wrong it has passed.

Overall resource consumption of the capacity has looked completely normal as well.  Whatever was wrong, its gone now and there is no trace of the issue.  The only commonality was timing.  Reports all seemed to succeed and fail around similar times of day.  The time of day was consistant, just when 1 report failed others around the same time also failed.  Later on in the same day they would succeed.

[V-lianl-msft]

Hi @Anonymous ,

I would suggest you to create a support ticket or submit your issue.

Best Regards,
Liang
If this post helps, then please consider Accept it as the solution to help the other members find it more quickly.

[Anonymous]

@V-lianl-msft  thats unncessary as the problem has resolved itself.