parallelize creating reports

[kri-2]

Hi,

I just wanted to know, if there is a way to parallelize the creation of reports. At the moment if I run 2 reports (that need a few minutes for completion) at the same time the first one is in "processing" state, the second one goes in "queued" state until the first finishes.

This leads to long runtimes e.g. if I want to create one report for 10 or more vms.

The Background for this question is that we use the reports via script which creates all reports first and then sends them per mail.

Any help would be appreciated.

Chris

[mark_j]

You can't edit the reporting engine to run reports in parallel vs series. Have you ever thought about scheduling the reports a certain time of the day, then have them dump to a network share, which your script ca just pick up and e-mail out in one message? This way your script would simply be picking up PDFs, attaching to an e-mail and sending, rather than messing with the vR Ops API.

[kri-2]

Hi Mark,

thanks for your answer, could you please give me some details, on how to edit the behaviour?

I thought about pre creating the reports but it does not work for me, since creating the report for one vm lasts ~4-5 minutes.

If I pre create the report for every vm I would need about 10 days :-).

So the report should only be created when really needed.

Chris

[mark_j]

Creating a report shouldn't take 4-5 minutes for a single VM. I'd seriously consider the time range you're reporting on of the report and try to limit  the period of time it's evaluating. I'd also suggest checking the health of your solution and ensure it's sized properly and have enough IO per the sizing guidance for your collection population.

[kri-2]

Hi Mark,

thanks again for your assesment. I also thought about the long running time of the report and if it is sane or if something is wrong. I hoped to get some time improvement with the update to vRops 6.2, but it is still the same.

What the report includes. And yes, it includes much - it should be a "all in one report" for our customers:

One VM:

last day cpu (usage, ready)

last day mem (usage, active, compressed, ballooned, swaped)

last day disk (latency write/read, usage)

last day network (EA in,out, lost)

last week cpu (usage, ready)

last week mem (usage, active, compressed, ballooned, swaped)

last week disk (latency write/read, usage)

last week network (EA in,out, lost)

last year cpu (usage, ready)

last year mem (usage, active, compressed, ballooned, swaped)

last year disk (latency write/read, usage)

last year network (EA in,out, lost)

I know, this is a big report, but I really can not believe that vRops needs 5 minutes to gather the metrics and data. Viewing this data in vCenter is done in view seconds (after choosing the according metrics and time range)...

What I am also worrying about is that our vRops only runs since 3 months now. So gathering the "last year" data could really get even longer.

What do you think about the duration of the Report? Are you still thinking it runs to long?

Chris

[sxnxr]

Chris If your report does not specify the vm and you select it when generating it can you export the report and up load it i will run in against a VM in our lab and see what the time is.

[kri-2]

Hi,

here is the exported report.

After importing it it can be run against a single vm.

Chris

[sxnxr]

I imported your report and ran it. It took 7 Mins to run against 1 VM. I modified it and creating an xls takes 5 seconds. Creating the PDF takes 6 mins. It must be the amount of rendering that takes place generating the views 12 chart views in the PDF. 

[kri-2]

This is really interesting, thanks for your effort.