Is PHP7 release stable?

[slawomir-pryczek]

Hi Guys,
I tried to switch to php-memcached on a service where we have like 20-30k request/s and it seems that PHP7 memcached release is having some issues... basically when using persistent connection mode it segfaults like crazy. I tried to switch to normal connections but then other memcached clients using old library under php5.6 were having issues to connect (maybe connections are not fully closed?)

We also made a simple debugging script, on pconnect it segfaults and loose key values (even if key is in cache it randomly returns empty), on normal connect it's much slower than memcache and unstable... connect->set->get takes about 1-30 ms, on the old setup (memcache/php) extension it's preety constant at 1-4ms.

Up to some point it works ok (we build a cluster like setup where we can control traffic send to all nodes), but after traffic gets higher it starts to make issues.

Can this be somehow related to connection handling and are you aware of this issue?

[mkoppanen]

Hello,

there is no release for PHP7 yet. I am currently working on resolving issues with PHP7 and cleaning up the code. Is the persistent connection issue easy to reproduce? I'll take a look as soon as possible

Thanks,
Mikko

[slawomir-pryczek]

Aha, that's great i'll see about reproducing the issue and prepare some code

[slawomir-pryczek]

Hi Mikko, we looked more into the issue...

Good thing is that the spike problem seems to be networking related. We have just installed php7/lighty on this server and it seems we have some switch problems. That's what was causing spikes in connection times... sorry for confusion.

Now for the segfault issues, we were able to eliminate it by switching to memcache (not d) module found here
https://github.com/websupport-sk/pecl-memcache/tree/NON_BLOCKING_IO_php7
Wasn't able to reproduce it in VM...

So the only thing i know for now ...

1. Segfaults disappear after switching memcache_d_ to normal connections
2. Segfaults also disappear after using memcache (without d) and persistent connections

If i'll find anything else that might help diagnosing issue - will post here.

Thanks.

[jonathanpmartins]

@mkoppanen Only waiting for stable release! My unit tests are passing!

[dzuelke]

Let me know if there's anything I can do to help with PHP 7, @mkoppanen !

[jonathanpmartins]

👍 ehehhe, oh, the waiting is killing me.

[mkoppanen]

@dzuelke,

Reviewing #214 would be helpful. Also running your local test suites against that would be a huge help. Been fixing a lot of errors in the existing code and the diff just keeps growing.

Among other things we had six versions (copy-paste) of different set and get code.

[mkoppanen]

@slawomir-pryczek,

with PR #214 I am unable to reproduce the persistent connection segfault. In my local environment persistent connections seem to work as expected and valgrind isn't complaining.

Can you try this branch and see if it solves your issues?

Thanks,
Mikko

[slawomir-pryczek]

Hi Mikko, sorry for waiting, we were able to do some quick tests with recent code (i saw #214 is merged in php7 branch), with persistent connections - no longer segfaulting, so it seems fixed.

[ralfbecker]

Any news or timeline for a PHP7 supporting release?

[jonathanpmartins]

@ralfbecker You can compile it using the php7 branch. See my comment on this issue: #194 (comment)
But I think its not stable yet. Good that my basic unit tests are all passing! hehehee

[oerdnj]

@mkoppanen JFTR I got one user reporting BC break in 3.0.0b1, but he refused to report it here and claims the BC break in 3.0.0b1 is deliberate. Are you aware of any BC breaks in 3.0.0b1 (latest git), I haven't found anything in the git log that would suggest so?

[mkoppanen]

@oerdnj, there is an intentional backwards compatibility break with cas token parameter in certain calls.

[dzuelke]

Is that documented anywhere? Or a ticket for it?