Since a couple of releases ago, the Desktop app has become unmanageable for my computer. It is an i5 with 8gb of RAM.
Upon inspecting what's going on, here's what i see.
Whenever the desktop app is opened, just the app, no report, it takes 1gb of RAM straight away.
Opening any report, and even having the option "Allow data preview to download in the background" disabled, power bi triggers three mashup evaluation containers that i don't know what are doing. The option is disabled, the report is idle, i have no power query window open... what are those mashup containers gathering?
Take a look at this screenshot, these are the processes that are triggered upon just opening a 10mb report, and doing nothing on the desktop, just staring at the main page. There's an element without name (1) that is taking 1gb of memory and three evaluation containers (2) taking 750mb of memory and around 75% of the CPU, i repeat, without doing anything on the desktop app:
I can leave the app open with the report idling and these three threads will not stop anytime, however the amount of CPU and RAM they use may vary. Eventually, the "no name" element, will display the report's name.
If i open another report we have, 80mb pbix, this is what happens:
This, just by opening the report, "Allow data preview to download in the background" is still disabled.
If i try to edit a query -which is being sourced from dataflows so the refresh and or the preview is usually very fast-, the preview takes ages to provide results. The same query, if it's closed and changes applied, will take seconds to retrieve the data. Why executing the whole query works as intended, but the preview which is basically getting the first 1000 results is not able to finish?
What i can see that may explain this behaviour, is that while working on power query and manually triggering a preview refresh, no new mashup engines are being triggered: the same three that were previously running keep on doing whatever they're doing, but no new ones appear. However, when i close and apply, there are indeed new mashup engine threads triggered on top of that three.
In other words, those three threads running right away since the file is opened, seem to be preventing the preview threads from running.
All this renders the power query interface useless. It is very hard to do anything there in these terms.
This report used to work just fine, queries could be edited and there was no drama to get the data in the preview. It usually takes ten minutes to refresh the whole report, and it takes around the same time when refreshed in the service.
Yesterday it took around an hour to have it refreshed from the desktop app, and i suppose it is because those three mashup engine threads were interfering in the normal report refresh.
Can anybody please shed some light around what these three mashup engine threads are doing or what is triggering them?
Is there anything that can be done to shut them down?
Thanks in advance for your help.
@v-jayw-msft , since the last update (2.85.985.0 64-bit sept 2020), where enhanced metadata is now generally available, this issue is striking back.
Before, we were able to disable the preview feature and the desktop app could be used normally.
Simply opening a report that loads no data (all queries have load disabled), takes up 4.5 gb of memory and 3 idling mashup evaluation containers are constantly open while there's no data load request specifically triggered by the user.
We should have more control about this. I need to be able to ensure that the desktop is focusing on what i need to do at the moment, and not whatever tasks it thinks it has to perform. And if these tasks are so necessary, at least inform the user about what's going on, and provide an expectation of when these tasks are going to finish.
Are there any investigations in progress as to why this is happening and when will it be solved?
Thanks in advance for your help.
I'm experiencing the same issue, with different types of reports, they worked flawlesly up until march version but from march version it started consuming absurd amounts of cpu just sitting idle on unnecesary mashup engine containers
For what can i see mashup container is trying to update data on the background event if not needed to, it's extremely frustrating because i cannot really work with any project as it makes my pc crawl due to the high cpu usge and temp on my machine
Happens both on x86 and x64 versions
It gets even triggered on a template load BEFORE you fill the parameters, so it should be not loading anything, just seems that the mashup containers are trying to retrieve and process the data before it's needed or even available.
i have just deployed a clean machine to azure just to test it and it works better... in that machine (it's only 4 cores and my machine is 4 x 2 with hyperthreading) and there it only spawns ONE mashupcontainer that uses cpu and it ends fast, less than a minute
We have narrowed down to enhanced metadata format migrated reports, the old ones do never trigger the issue.
Also i have found that if i clear the credentials to data source access the mashup container does not spawn until i set them again, but at that very moment they start again, it even locks the credentials dialog of the datasource to end until the mashup container is killed, it seems that is waiting for its evaluation to end
Please check the offical document about the minimum requirements of Power BI Desktop.
And refer to the topic below about Microsoft mashup evaluation container.
Hi @v-jayw-msft ,
Thanks for your reply.
Regarding the minimum requirements, i'd say the computer i'm using is covering them.
Regarding the link about mashup engine you're providing, i apologize in advance, but i fail to see the connection with the issue discussed in this thread.
Just to summarize, the same set of queries, placed in two different pbix files with the same settings, are providing different behaviour: One file is triggering three mashup engine threads upon opening and the other one is not.
There may be a hidden setting in one of the files that is causing this, or there may be another explanation.
The point being that these three mashup engine threads have not been triggered on purpose by any action made by me, such as refreshing the model or editing a query through the transform data UI, cannot be killed through task manager, and are "sucking" precious computer resources which are degrading the overall interaction with the app, making any edit anywhere a pain.
The purpose of this post is to learn what's going on, why is it occurring and how can we prevent it from happening.
Thanks and kind regards,
I've made a quick test, I created a new pbix and copied over all the queries from the other report. The behaviour has been different, and this is what task manager displays upon opening the file:
As you can see, there are no mashup engines triggered when the file is opened.
This means that there's something within the original file that is somehow triggering those mashup engines upon file opening.
Anybody has an idea about how to track the source of the trigger?
Watch the playback when Priya Sathy and Charles Webb discuss Datamarts! Kelly also shares Power BI Community updates.
Watch Session 24 of Ted's Dev Camp along with past sessions!
The first Microsoft-sponsored Power Platform Conference is coming in September. 100+ speakers, 150+ sessions, and what's new and next for Power Platform.