Reduce parallelism for Assets imports

This recommendation is shown when we’ve detected that Assets imports are running with high parallelism settings and are likely contributing to high CPU usage and slower page response times.

Signals used

Details of how we detect this issue.

Conditions

The following conditions need to occur together:

  • At least one key experience, such as viewing or editing issues, shows degraded page load times (RFU).

  • During the same time window, one or more Assets imports were running.

  • One or more of those imports were CPU‑bound, meaning most of their total runtime was spent doing CPU work rather than waiting on I/O.

  • CPU usage is high and the nodes running these imports are saturated.

  • The configured Assets import parallelism settings are high.

When these conditions are met together, we surface this recommendation and suggest lowering the parallelism settings.

How we detect it

We use a combination of Assets events, configuration, and resource usage:

Signals

Details

RFU degradation

We look for Ready for User (RFU) degradation - cases where key Jira experiences take significantly longer than their historical baseline. This helps us confirm that users are actually experiencing slower page loads, not just one-off spike.

Assets import completed events

We use Assets import events to see whether imports were running during the RFU degradation window. From these events, we determine whether the imports were CPU‑bound and how long they ran.

Assets configuration

We read the current Assets configuration to get:

  • Assets parallelism - the number of threads each import can use

  • Max parallel imports - how many imports can run at the same time

If these values are above conservative defaults (more than 2 threads per import, more than 1 import in parallel), we treat the parallelism as high.

CPU and memory usage

We look at node‑level JVM CPU metrics to see whether the nodes running imports were saturated while imports were executing. We also collect memory usage for context, but it does not directly trigger this recommendation.

Impact of imports vs. no imports

We compare RFU in the degraded window to a similar time window (up to six weeks back) where fewer imports were running. If RFU was better when fewer imports were running, we increase our confidence that high parallelism is contributing to the issue and may raise the priority of the recommendation.

This combination helps us focus on situations where parallelism settings are likely amplifying CPU usage and affecting performance.

Filters

When the issue occurs multiple times, we group occurrences under different time filters, for example a 24‑hour or 7‑day period. In this case:

  • For degraded experiences, we show the highest degradation from all occurrences in the selected time filter.

  • For configuration details, we show the highest setting values observed across all occurrences. These usually stay the same unless you changed the configuration between occurrences.

This helps you see whether the problem is short‑lived or persistent, and which imports are most often involved.

Frequency chart

The frequency chart shows how often this issue has occurred over time. This can help you identify patterns such as:

  • times of day or days of the week when heavily parallel imports cause this issue

  • whether this issue was caused by past changes to import schedules or parallelism settings

How you address this issue will depend on your import workload and when imports usually run. Here are some actions you can take:

Improve the import configuration

Lower the Assets parallelism and reduce the number of parallel imports. You can configure both by going to Jira administration > Assets configuration.

Assets parallelism

This is the number of threads per import. Reducing it can lower CPU pressure on nodes during heavy imports, especially when imports overlap with peak user activity. You can lower this setting incrementally and monitor for improvements.

Max parallel imports

This is the number of imports that can run at the same time. Reducing it can help avoid multiple CPU-heavy imports using the same resources on the same node.

For details, see Configuring parallel imports.

Monitor for improvements

After adjusting your import configuration, monitor RFU during the next few import runs. If the recommendation stops appearing, it’s a good sign that your new settings are a better fit for your environment.

Still need help?

The Atlassian Community is here for you.