[ https://issues.apache.org/jira/browse/TS-2604?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13922413#comment-13922413 ]
Ethan Lai commented on TS-2604: ------------------------------- [~zwoop] Thanks a lot for reviewing this plugin, It's ok to align 2 of this plugin's configurations (insert_lock_retry_time & max_lock_retry_timeout) to use the 2 overridable configurations you mentioned. However, 1. It's too late to set these 2 overridable configurations in CACHE_LOOKUP_COMPLETE hook, the cache lookup process is done and it's unable to make this txn to lookup again in current version. 2. If I use TSCacheRead() to check cache object's exist or not, I still don't know it's fresh or stale. (MISS & STALE perform the same way in this plugin) 3. For uncacheable requests, it would impose big latency as you mentioned. In my use cases, it's not easy to determine the response of a request will be cacheable or not from it's uri nor headers, maybe you have good ideas in distinguishing them. :) I chose to hook in POST_REMAP is because it's the last hook before cache lookup. I can pause requests with same CacheUrl to proceed its cache lookup process till response of previous request is received. Then these followup requests might be able to take advantage from cache HIT_FRESH. > Collapsed connection plugin > --------------------------- > > Key: TS-2604 > URL: https://issues.apache.org/jira/browse/TS-2604 > Project: Traffic Server > Issue Type: New Feature > Components: Plugins > Reporter: Ethan Lai > Assignee: Bryan Call > Labels: review > Fix For: 5.0.0 > > > This plugin collapses connections with identical CacheUrl/EffectiveUrl. > Please find more detail from README and state.png (State Diagram) -- This message was sent by Atlassian JIRA (v6.2#6252)