Reduce reindexing thread count

This recommendation appears when the configured indexing thread count is significantly higher than the CPU capacity. Confluence can use multiple threads to parallelize indexing work, but setting the thread count too high can have the opposite effect. Instead of improving throughput, too many indexing threads can overload the application node, increase context switching, compete with user-facing requests, and create unnecessary load on the database. This may result in slower page loads, slower editor startup, delayed publish actions, or degraded dashboard and search responsiveness while the reindex is running.

Signals used

The following conditions need to occur together for us to detect this issue:

• index.queue.thread.count or reindex.thread.count is significantly greater than the number of CPU cores on the node.

• The configured value is four times or more than the available CPU capacity.

This recommendation can also be shown if no ready-for-user (RFU) time degradation is currently observed, because the thread count configuration is a known risk and can be challenging to diagnose later.

フィルター

We group their signals under different filters, for example, Last 24 hours or Last 7 days.

• In the Degraded experiences section of the recommendation, we show the highest degradation across all occurrences for experiences such as viewing, creating, editing, or publishing a page, dashboard rendering, and using the search function.

• In the Reindexing thread count section, we list the number of reindexing threads, nodes over limit, versus the available CPU capacity.

頻度

Similar to the filters, the Frequency section shows the impact of high thread count over time. It can help you spot specific patterns.

推奨されるアクション

Here are some actions you can take to resolve this issue:

Lower the indexing thread count

Review the configured values of index.queue.thread.count and reindex.thread.count. If either is set far above available CPU, reduce it to a more conservative value. The minimum allowed value is 2. Reducing indexing thread count can lower CPU pressure, reduce database stress, and make reindexing more predictable. In some cases, a lower setting results in a faster overall completion time because the system spends less effort managing excessive parallel work.

Additionally, we recommend you avoid pushing indexing parallelism well beyond what the node can support. For example, if a node has 8 CPU cores and the thread count is set to 50, the setting is likely too aggressive and should be reduced substantially.

Schedule reindexing for off-peak hours

If reindexing is necessary, run it during lower-traffic periods whenever possible. This reduces the chance that indexing work will compete with normal user activity, causing noticeable slowdowns.

Consider isolating background jobs

For larger environments, consider using a dedicated node for background or maintenance-heavy work. Isolating reindex load can help protect user-facing traffic from spikes in CPU and database usage.

Reassess after the change

After lowering the thread count, monitor reindex duration, CPU utilization, queue behavior, and end-user performance. The goal is not only to reduce load, but also to ensure reindexing completes steadily without stalling.