"az consumption usage list" command cannot be run in parallel

[zenfish]

Related command

az consumption usage list

Describe the bug

The command can't be parallelized. You essentially have to run it serially (well, sometimes I can do 2, maybe 3 at once... woohoo!) Combined with the very long execution time (#23204), this means I can't run daily reports. All other az-cli commands support at least a tiny bit of parallelization, most easily handling fifty or more jobs at a time.

Running this sub command in parallel (20 at a time) over our 1059 subscriptions returned:

63 successful
23 Returned no output (perhaps succeeded, perhaps not, haven't investigated)
973 failures

Failure breakdown according to the ERROR messages given by az-cli -

926 ERROR: (429) Too many requests. Please retry after 60 seconds.
 44 ERROR: (SubscriptionNotFound)
  2 ERROR: (401) Unauthorized
  1 ERROR: (400) Subscription scope usage is not supported for current api version.

The 44 that weren't found... a bit odd, since this was run immediately after running az account list
to get the subscriptions in the tenant.

On the bright side the failed run only took a bit over 4 minutes to complete.

To Reproduce

Here I'm using the rush command to parallelize shell commands, as I do with all the other az-cli sub commands I use (rush is at https://github.com/shenwei356/rush); you can run from 1-N jobs in parallel at a time; when it finishes one job it starts another. Here rush is reading from standard input... the file "subs.txt", which has, one per line, the IDs of the > 1000 subscriptions we run in this tenant.)

First, the simple "costs.sh" shell script that will be run in parallel; it puts the result in /tmp -

start=$(gdate -d "last month" "+%Y-%m-01")
end=$(gdate -d "$(date +%Y/%m/01) - 1 day 00:00" +"%Y-%m-%d")
date_filter="-s $start -e $end"

az consumption usage list -a -m $date_filter --subscription "$1" > /tmp/$1.res

The output of running the above script under rush is attached (gitout.txt) the -j option below specifies the number of jobs to run in parallel (20 here.)

(see attachment)

gitout.txt

Expected behavior

The command should be able to be run in parallel from a single endpoint and return the results within a reasonable time frame. The cost web UI, which has the costs sliced and diced, actually loads pretty quickly.

Environment summary

Installed via brew on a macbook and iMac, currently running MacOS 10.16 and/or 10.15.7, but this has been slow for literally the last few years no matter what the OS or az-cli versions. Here's the macbook version, which was used for the above test and output -

`az version
{
"azure-cli": "2.23.0",
"azure-cli-core": "2.23.0",
"azure-cli-telemetry": "1.0.6",
"extensions": {
"azure-devops": "0.18.0",
"databricks": "0.7.2",
"kusto": "0.2.0",
"log-analytics-solution": "0.1.1",
"resource-graph": "1.1.0",
"storage-preview": "0.7.2"
}
}

**Additional context**

<!--- Add any other context about the problem here. --->


The `consumption` sub command has been in preview for years. Come on. From the consumptio:
"WARNING: Command group 'consumption' is in preview and under development."

I can send a full debug log of a run (about 500-800MB) if you want (I can point to a URI or whatever, if you mail me.)

Thank you for your feedback. This has been routed to the support team for assistance.

[yonzhan]

route to CXP team