-
Bug
-
Resolution: Fixed
-
Highest
-
6.6.6, 6.13.0, 6.13.4
-
33
-
Severity 1 - Critical
-
71
-
Summary
The content API endpoint is available to retrieve content from a Confluence instance. While most users pass arguments that help to filter down to specific pieces of content, it is possible to hit the generic endpoint which fetches the entire contents of a given Confluence instance. When this happens, the threads that are tasked to run the operation can run longer than expected, especially with larger Confluence instances. In addition, this is especially true for calls to this endpoint that are made anonymously (as the endpoint contains the @AnonymousAllowed designation) and a Confluence instance is not set up for anonymous access. When this occurs, the operation can still run for a longer than expected time and may ultimately return an empty set to the user.
Environment
- Confluence
Steps to Reproduce
- Fill a Confluence with several hundred thousand -> million pages
- run the following command curl -X GET http://<baseURL>/rest/api/content | python -mjson.tool
Expected Results
- the operation will run and will not impact performance on the instance
Actual Results
- the operation can be expensive and cause performance to degrade
Workaround
Confluence 7.18 and later
Use http://<baseURL>/rest/api/content/scan instead. Examples how to use endpoint could be found at the page Confluence REST API examples
Older versions
Prevent calls to the base end point via networking/proxy rules
- mentioned in
-
Page Failed to load
-
Page Failed to load
-
Page Failed to load
-
Page Failed to load
-
Page Failed to load
-
Page Loading...
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
| Form Name | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
[CONFSERVER-57639] Hitting the base Content API endpoint can cause performance problem for large instances
| Remote Link | Original: This issue links to "Page (Confluence)" [ 643508 ] |
| Remote Link | New: This issue links to "Page (Extranet)" [ 713569 ] |
| Remote Link | New: This issue links to "Page (Extranet)" [ 713483 ] |
| Resolution | New: Fixed [ 1 ] | |
| Status | Original: Waiting for Release [ 12075 ] | New: Closed [ 6 ] |
| Remote Link | New: This issue links to "Page (Confluence)" [ 645814 ] |
| UIS | Original: 47 | New: 71 |
| UIS | Original: 23 | New: 47 |
| Support reference count | Original: 32 | New: 33 |
| Description |
Original:
h3. Summary
The content API endpoint is available to retrieve content from a Confluence instance. While most users pass arguments that help to filter down to specific pieces of content, it is possible to hit the generic endpoint which fetches the entire contents of a given Confluence instance. When this happens, the threads that are tasked to run the operation can run longer than expected, especially with larger Confluence instances. In addition, this is especially true for calls to this endpoint that are made anonymously (as the endpoint contains the @AnonymousAllowed designation) and a Confluence instance is not set up for anonymous access. When this occurs, the operation can still run for a longer than expected time and may ultimately return an empty set to the user. h3. Environment * Confluence h3. Steps to Reproduce # Fill a Confluence with several hundred thousand -> million pages # run the following command *{{curl -X GET http://<baseURL>/rest/api/content | python -mjson.tool}}* h3. Expected Results # the operation will run and will not impact performance on the instance h3. Actual Results # the operation can be expensive and cause performance to degrade h3. Workaround h4. Confluence 7.18 and later Use *{{http://<baseURL>/rest/api/content/scan}}* instead. See examples how to use h4. Older versions Prevent calls to the base end point via networking/proxy rules |
New:
h3. Summary
The content API endpoint is available to retrieve content from a Confluence instance. While most users pass arguments that help to filter down to specific pieces of content, it is possible to hit the generic endpoint which fetches the entire contents of a given Confluence instance. When this happens, the threads that are tasked to run the operation can run longer than expected, especially with larger Confluence instances. In addition, this is especially true for calls to this endpoint that are made anonymously (as the endpoint contains the @AnonymousAllowed designation) and a Confluence instance is not set up for anonymous access. When this occurs, the operation can still run for a longer than expected time and may ultimately return an empty set to the user. h3. Environment * Confluence h3. Steps to Reproduce # Fill a Confluence with several hundred thousand -> million pages # run the following command *{{curl -X GET http://<baseURL>/rest/api/content | python -mjson.tool}}* h3. Expected Results # the operation will run and will not impact performance on the instance h3. Actual Results # the operation can be expensive and cause performance to degrade h3. Workaround h4. Confluence 7.18 and later Use *{{http://<baseURL>/rest/api/content/scan}}* instead. Examples how to use endpoint could be found at the page [Confluence REST API examples|https://developer.atlassian.com/server/confluence/confluence-rest-api-examples/] h4. Older versions Prevent calls to the base end point via networking/proxy rules |
A fix for this issue is available in Confluence Server and Data Center 7.18.0.
Upgrade now or check out the Release Notes to see what other issues are resolved.