You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Following up on the memory-growth report on 2.8.24.
#5062 (against 2.8.x) and #5064 (against 2.9.x) were opened for this issue, then closed on 6 March 2026 without being merged, and their head branches were deleted. I compared SwaggerLoader.java in 2.8.24 with the later published 2.8.31 and 2.9.5 releases; the file content is the same, so I do not see this change in a published release.
On #5064 there was also an open question about whether the proposed cache-key change would actually address leaks caused by frequent version upgrades. I have not verified a root cause or a working fix.
Could you confirm whether you can still reproduce this on a later published 2.8.x or 2.9.x release, and share sanitized memory / load schema from service center observations if you can? That will help us decide the next investigation step.
Activity
chengyouling commentedon Jan 20, 2026
应该不会是这里的问题,可以观察下是不是一直打印load schema from service center......相关日志
chengyouling commentedon Jan 20, 2026
上游服务A长时间不重启,下游服务B频繁变更升级,导致A不停加载B服务契约,导致内存消耗增加。
[apache#5061] fixed service upgrade caused memory leak problem
[apache#5061] fixed service upgrade caused memory leak problem.
[apache#5061] fixed service upgrade caused memory leak problem.
hezhangjian commentedon Nov 13, 2026
Following up on the memory-growth report on 2.8.24.
#5062 (against
2.8.x) and #5064 (against2.9.x) were opened for this issue, then closed on 6 March 2026 without being merged, and their head branches were deleted. I comparedSwaggerLoader.javain 2.8.24 with the later published 2.8.31 and 2.9.5 releases; the file content is the same, so I do not see this change in a published release.On #5064 there was also an open question about whether the proposed cache-key change would actually address leaks caused by frequent version upgrades. I have not verified a root cause or a working fix.
Could you confirm whether you can still reproduce this on a later published 2.8.x or 2.9.x release, and share sanitized memory /
load schema from service centerobservations if you can? That will help us decide the next investigation step.