HandlerSocket on Percona 5.6

[ecommunities]

I was a little dismayed after spending several hours updating my chef files and dealing with deprecated features and finally getting everything up and running, only to learn that HandlerSocket hasn’t been implemented in 5.6. I missed that note in my first glance at the “Changes”

I’ve only found minimal reference to it, a couple questions on SO, and here, but with no answers. The only semblance of a response referenced it not being compatible and that you may “consider” it. I can only assume that that was during the Alphas of Percona 5.6 since I’ve looked around at DeNA, and they’re currently supporting 5.6.12.

It’s a major component of our service, without it there’s no chance of us upgrading with Percona. I’ve been patiently waiting for the fulltext capabilities of 5.6, but it can’t be at the expense of “nosql”.

I was hoping someone could clear the air and provide a more firm answer as to whether Percona will be re-including HandlerSocket.

Thanks!

[laurynas]

Yes, we are looking to include HandlerSocket with Percona Server 5.6.

The reason why it was not present in the alphas/RCs was that 5.6-compatible HandlerSocket was not ready at the time. I have just checked and I see that 5.6 support has been merged to the main HandlerSocket tree, making it possible for us to merge it. I cannot promise though that it will be the 1st GA release that will contain it, it could be a point release afterwards.

[ecommunities]

That’s great, fingers crossed on the GA, but as long as I know it’s on the radar I can bare the wait Cheers, and thanks for producing a top notch product!

[ecommunities]

@laurynas, is there any update on the availability of Handlersocket in 5.6? I participated in one of the recent webinars and asked the question but they never got to it and I haven’t seen anything new. Any update would be appreciated, thanks!

[andrisi]

hello @laurynas! any news on this? we would also be very glad to switch to percona server 5.6 if this was available. thanks!

[ecommunities]

Bump. What’s up guys, you’re at point release 16 and still no word of handlersocket.

Tom mentioned in Stewart’s blog a while back that you were aiming for Q1 2014 that’s now come and gone, can someone please let us know what your intentions are, if you’re going to drop it can you just come out and say that, it’s getting a little frustrating when you keep saying it’s coming and leave us all hanging.

[Ringdingcoder]

Having heard about the InnoDB memcached plugin today (from Oracle), I am under the impression that that’s what’s supposed to be used going forward. It seems to be similar enough in functionality to HandlerSocket to act as a replacement in many cases.

[ecommunities]

@Ringdingcoder The memcache plugin is still not as performant as the handlersocket plugin and for those of us that dove in, because someone like Percona was supporting it, despite having the DB layer abstracted, it would still be a major task to rewrite around memcache. Handlersocket was built specifically for this purpose, while memcache is being bastardized to do so. Handlersocket is just plain superior, and Percona’s been saying for a year they were going to release it. Time to pony up.

Even if we only got 5.6 with handlersocket and it was officially dropped going forward, we’d be aware of that move and be able to plan for it. Percona dropping support in 5.6 temporarily was a surprise which was more than a little disappointing. It would also give memcached some time to mature in the MySQL branch and stabilize, maybe even reach the same level of performance as handlersocket.

[andrisi]

Just read that HandlerSocket is included in 5.6.17-66. Thanks a lot!

[ecommunities]

Too bad that announcement didn’t make it into one of the release newsletters, don’t think I missed that?!?! But all the same, THANK YOU!!!