Reporting fails on large dashboards

[liza-mae]

Kibana version: 7.13.0 BC9

Elasticsearch version: 7.13.0 BC9

Server OS version: Ubuntu

Browser version: Chrome latest

Original install method (e.g. download page, yum, from source, etc.):
Staging

Description of the problem including expected versus actual behavior:
Reports error on large dashboard.

Steps to reproduce:

1. Install ES/Kibana
2. Generate report on a large dashboard
3. ecommerce sample data + 60 more visualizations (manually) and the functional test fails

Provide logs and/or server output (if relevant):

{"type":"log","@timestamp":"2021-05-24T13:51:22+00:00","tags":["info","plugins","reporting","chromium-stderr"],"pid":18555,"message":"[0524/135122.246386:ERROR:gpu_init.cc(426)] Passthrough is not supported, GL is swiftshader"}

{"type":"log","@timestamp":"2021-05-24T13:49:47+00:00","tags":["error","plugins","reporting","printable_pdf","execute-job","kp2o13no0ebf8f5fe5fcpubj"],"pid":18555,"message":"Error: Protocol error (Page.captureScreenshot): Unable to capture screenshot\n    at /home/vagrant/kibana-7.13.0-linux-x86_64/node_modules/puppeteer/lib/cjs/puppeteer/common/Connection.js:208:63\n    at new Promise (<anonymous>)\n    at CDPSession.send (/home/vagrant/kibana-7.13.0-linux-x86_64/node_modules/puppeteer/lib/cjs/puppeteer/common/Connection.js:207:16)\n    at Page._screenshotTask (/home/vagrant/kibana-7.13.0-linux-x86_64/node_modules/puppeteer/lib/cjs/puppeteer/common/Page.js:1105:43)\n    at runMicrotasks (<anonymous>)\n    at processTicksAndRejections (internal/process/task_queues.js:93:5)"}

[elasticmachine]

Pinging @elastic/kibana-app-services (Team:AppServices)

[elasticmachine]

Pinging @elastic/kibana-reporting-services (Team:Reporting Services)

[liza-mae]

I labelled this a regression since it appears to be worse than 7.12.1 reporting and related to this recent PR: #99853

[LeeDr]

The test is https://github.com/elastic/kibana/blob/7.13/x-pack/test/functional/apps/dashboard/reporting/screenshots.ts

These tests are not failing on Kibana CI but looks like they started failing immediately on Cloud when the PR was merged and backported.

[liza-mae]

I suspect it passes on Kibana CI because the workers have a higher hardware specification. I ran a very large dashboard as noted above on same hardware specification for 7.12.1 and 7.13.0 and it failed on 7.13.0, I used 2GB - 4GB RAM, have not tried higher yet.

[liza-mae]

I confirmed the test passes locally on 7.13.0 with 8GB of RAM, need @tsullivan to confirm if this the new requirement.

[tsullivan]

I confirmed the test passes locally on 7.13.0 with 8GB of RAM, need @tsullivan to confirm if this the new requirement.

It turns out we can't predict how much RAM is needed to generate a report of a particular dashboard, even if we know the height at the time the user requests a report.

For environments where Reporting is a must-have, the guidance is to give the machine as much RAM possible. I would start at 16GB and go down from there, to find the instance size that is too small for the report.

I understand this isn't a long-term solution: there should not be a trial-and-error setup. We are working on a solve this in a good, cost-effective way.

[vadimkibana]

Thank you for contributing to this issue, however, we are closing this issue due to inactivity as part of a backlog grooming effort. If you believe this feature/bug should still be considered, please reopen with a comment.

---

We are aware that large report generation consumes a lot of memory on Kibana instances and sometimes causes failure of report generation process.

• In general, we recommend using 8GB+ memory instances, if report generation will be used.
• Over the last year a number of improvements have been made to use less memory.
• We have few more memory improvements on our roadmap.