Let's summarize here all critical issues to make stable release
This related both for DRUPAL-6 and DRUPAL-5 branches
After spending a lot of time reading through current issues I realized
1) file-engine is broken in current RC1 so lets focus on
#578522: File engine: Performance, filenames, and other cleanup
#588820: Creating severe load on system by logging thousands of unlink errors in DB log
2) There's a some class conflicts #564460: Class conflict, namespace collisions
We should take D7 approach with class naming to prevent this
3) There's a big trouble with locking in APC and memcache which have no own
- Memcache should be synced with memcache module or rolled back to beta9
- It's a good idea to try implement php-semaphores http://php.net/manual/en/book.sem.php or any available on
4) We need some UI #448274: Cache router wont show up in the Administration menu
- for 1.0 version seems enough < ahref="http://api.drupal.org/api/function/hook_requirements">hook_requirements()
- do we really need statistics?
5) there's a lot of documentation issues - configuration of this module is too hard or bad described
My English is not good to write docs - Any help is welcome
Comments
Comment #1
andypostWe need to fix tests for drupal6 #698276: Tests for module
For Release I'm planing to implement #463494: Add support for Zend Cache backend
there's a lot of info in #245341: Unthrottled while loop in memcache engine causing high cpu utilization.
Drupal 7 #593238: Do not release Drupal 7 version of this and move the engines into separate projects
Comment #2
slantview CreditAttribution: slantview commentedUpdates:
1) #588820: Creating severe load on system by logging thousands of unlink errors in DB log - This issue has to do with #3, NOT with file engine specifically. There are troubles with file engine, but I think we need to look at how we are implementing this and how we should change it. It was originally copied from fastpath_fscache module. That module looks abandoned. We can re-write some parts of our current file caching make it work well. I am very much ok with the first patch though I haven't looked at it much.
2) I'm totally ok with this as well. I would say let's using the naming convention from D7 as well and change it for all of the branches.
3) I think the current implementation of locking in Memcache is ok with the exception of figuring out how to do a versioning and not waiting on the lock for so long. APC and the rest of the opcode/memory caches should be reverted to beta9 version and worked out from there (just add back the lock/unlock functions into the classes). I think PHP semaphores are a little overkill however. I would say we should just figure out a consistent strategy to have atomic locks, that is why i tried out using file locking, but for the long run, i would say lock with whatever the technology is, or go back to using the database for that (see lock.inc for D7).
4) Yeah, this is so highly debatable. Because we can't consistently get stats from all of the backends, I would say we go with the stats from the least common denominator and just get at least some basic stats. There should be something. This should be version 2.0 however.
5) Since you are working on getting some of this fixed, I can work on documentation. Let me know what kind of documentation we should have. I can fix the readme file and figure out how to get some of this stuff documented.
As far as Zend Cache I'm not too worried about a Zend backend for now, and I almost feel like we should wait for a version 2.0 for that. I don't want Zend to hold up a 1.0 release for this. We can start a 2.0 branch as soon as we get 1.0 out.
Comment #3
andypostFirst of all we need to roll drupal-5 rc1 version looks like you forget to roll this when 6rc1 was done
I will commit class changes for 2) and I think it's time for next RC
Comment #4
slantview CreditAttribution: slantview commentedI would prefer to revert APC/XCache/eacc to revert to previous locking for rc2. If need be, I can make those changes.
Comment #5
andypostPlease try to revert, I'm working on tests most of engines have exceptions with lifetime checking...
Comment #6
andypostToday #251792: Implement a locking framework for long operations come to drupal-6 core. Should we use this? Suppose no, because it's too slow.
And we should deside on #564460: Class conflict, namespace collisions
Comment #7
andypostToday 5 branch was updated and I've made dev snapshot so waiting for feedback.
Comment #8
andypostNeed docs about eaccelerator 0.9.6 is not using cache anymore #576640: Support for eaccelerator
Comment #9
andypostalso need to change at all branches #710630: Memcached engine passes too many parameters
Comment #10
andypost@slantview we should decide on
#564460: Class conflict, namespace collisions
#588820: Creating severe load on system by logging thousands of unlink errors in DB log
I think a wincache is ready go #743028: Support for wincache
fix for docs #766748: Caution when running w/APC older than 3.1
and then push rc2
#245341: Unthrottled while loop in memcache engine causing high cpu utilization.
should go into next release or branch with different approach on cache clearing like #493448: Memcache and cache_clear_all wildcard
Comment #11
andypostFor D7 #800014: Tiny bootstrap clean-up
Comment #12
andypostFor D7 headers field removed from cache tables due #692044: A site with statistics module enabled is much slower in serving cached pages in D7 than in D6
Comment #13
andypost#588820: Creating severe load on system by logging thousands of unlink errors in DB log
#245341: Unthrottled while loop in memcache engine causing high cpu utilization.
currently $cache_router for #564460: Class conflict, namespace collisions
I think we should push another RC