Configuring parallel imports
You can run multiple imports simultaneously on your single-node or multi-node Data Center instance to accelerate Assets data imports. As these imports can be scheduled or started manually, you can initiate a manual import even when other scheduled imports are running. You can also be assured that scheduled imports run on time, even if an earlier import is taking longer than expected.
If you're using a dedicated node, all your imports will run on the same node.
Adjusting the number of parallel imports
By default, Jira Service Management lets you run up to five parallel imports across your cluster. You can adjust this number based on the capacity of your cluster or even choose not to allow imports to run in parallel.
Note that:
- this number can never be higher than the number for the Assets Parallelism setting to ensure your system runs smoothly.
- any time you adjust this setting, you won’t need to restart Jira, and your changes won’t affect the imports that are already in progress.
To configure this setting:
- Log in as a Jira admin.
- Navigate to Administration > Manage apps > Assets configuration.
- In the General configuration section, edit the number set in Maximum parallel imports.
Refer to the table below to understand some of the common scenarios where you might want to adjust the parallel imports setting.
| Action | Use case | Impact |
|---|---|---|
Increase parallel imports Note: Increase this number with caution, as imports are memory intensive. | You have enough system memory to run more imports than the number you’ve currently set for parallel imports so that you don’t have to wait for other imports to finish before they can start theirs. | Additional import processes need more system memory. Scheduled and manual imports are likely to run on time. |
Decrease parallel imports | You want to use system memory for other processes, or you don’t have enough memory to support multiple imports. | Import processes need less system memory. Scheduled and manual imports are likely to have delayed starts. |
Turn off parallel imports Note: To turn of parallel imports, set Maximum parallel imports to 1. | You might have multiple large imports that consume a lot of memory and don’t want them to run together, or you might want to run your imports in a specific order. For example, if you run an import on an object schema after your first import without parallel imports running, you are assured that the second import won’t start until the first one completes. | Scheduled and manual imports may not run on time. Import processes use the same amount of system memory as they did before parallel imports were available. |
Import progress and object locking
You will notice some additional steps during the import process only if the system is waiting:
- for other imports to complete
- to lock object types that are currently locked by other imports
These additional steps give you an idea of the number of imports due to complete before your import can start, or the number and progress of object types being locked.
Note that during parallel imports, the system locks relevant object types (listed below) across the cluster to improve data consistency:
- The object types configured in the current import.
- The object types referenced by the object types configured in the current import (including references from other schemas).
- All child object types referenced by the object types configured in the current import (including references from other schemas).
Memory and performance
Note that if all the parallel imports coincidentally end up on the same node, the imports will run slower as they still share the same max number of CPU threads allocated in the Assets Parallelism setting; however, they’ll be fairly balanced across all the imports. It is also worth noting that there is negligible overhead for this balancing, so the total runtime of multiple imports sequentially and parallel is the same.
Troubleshooting
If you run out of memory during multiple large imports, we recommend that you disable the imports, and then either increase the memory or turn off parallel imports before re-enabling the imports.