Subject: Re: Who's in the mood for a new memcached release? - msg#00039

Previous Message by Date: click to view message preview

Re: Who's in the mood for a new memcached release?

What did you have to leave out due to conflicts? -Steve Brad Fitzpatrick wrote: This is a merge of facebook's changes into the main version, with some of the new features that were conflicting removed, others partially still there, etc, etc. WARNING: we're not using it in production. I can't vouch for its quality. Facebook can vouch for the version they used, but maybe I messed it up merging it in. HERE BE DRAGONS. Be scared. That said, http://www.danga.com/memcached/dist/experimental/memcached-1.2.0-rc1.tar.gz Notice the /experimental/ in the URL? PACKAGERS: Don't package this yet. It's a dev release. Everybody else: bang on it. Let's find any issues and get a release out. It's been too long, and I want to finish vbuckets, but want a stable tree to hack on. Also wanted: benchmarks between this and last stable release. Feedback requested.... - Brad

Next Message by Date: click to view message preview

Re: Who's in the mood for a new memcached release?

I notice that this release doesn't have our powers-of-N allocation change -- was there something wrong with it such that it wasn't appropriate to include in the release? It gives us substantially more efficient memory usage and doesn't really have a downside that I'm aware of (but maybe there's one I'm not aware of?) -Steve Brad Fitzpatrick wrote: There was a "bget" command (binary get) in trunk that had to be removed. It's still in the 1.2.0-rc1, kinda, in that you can do "bget" which sets the binary flag, but the getter code doesn't do it now. We were doing that so C clients could be more efficient (and easier/safer to write). - Brad On Thu, 17 Aug 2006, Steven Grimm wrote: What did you have to leave out due to conflicts? -Steve Brad Fitzpatrick wrote: This is a merge of facebook's changes into the main version, with some of the new features that were conflicting removed, others partially still there, etc, etc. WARNING: we're not using it in production. I can't vouch for its quality. Facebook can vouch for the version they used, but maybe I messed it up merging it in. HERE BE DRAGONS. Be scared. That said, http://www.danga.com/memcached/dist/experimental/memcached-1.2.0-rc1.tar.gz Notice the /experimental/ in the URL? PACKAGERS: Don't package this yet. It's a dev release. Everybody else: bang on it. Let's find any issues and get a release out. It's been too long, and I want to finish vbuckets, but want a stable tree to hack on. Also wanted: benchmarks between this and last stable release. Feedback requested.... - Brad

Previous Message by Thread: click to view message preview

Re: Who's in the mood for a new memcached release?

What did you have to leave out due to conflicts? -Steve Brad Fitzpatrick wrote: This is a merge of facebook's changes into the main version, with some of the new features that were conflicting removed, others partially still there, etc, etc. WARNING: we're not using it in production. I can't vouch for its quality. Facebook can vouch for the version they used, but maybe I messed it up merging it in. HERE BE DRAGONS. Be scared. That said, http://www.danga.com/memcached/dist/experimental/memcached-1.2.0-rc1.tar.gz Notice the /experimental/ in the URL? PACKAGERS: Don't package this yet. It's a dev release. Everybody else: bang on it. Let's find any issues and get a release out. It's been too long, and I want to finish vbuckets, but want a stable tree to hack on. Also wanted: benchmarks between this and last stable release. Feedback requested.... - Brad

Next Message by Thread: click to view message preview

Re: Who's in the mood for a new memcached release?

I notice that this release doesn't have our powers-of-N allocation change -- was there something wrong with it such that it wasn't appropriate to include in the release? It gives us substantially more efficient memory usage and doesn't really have a downside that I'm aware of (but maybe there's one I'm not aware of?) -Steve Brad Fitzpatrick wrote: There was a "bget" command (binary get) in trunk that had to be removed. It's still in the 1.2.0-rc1, kinda, in that you can do "bget" which sets the binary flag, but the getter code doesn't do it now. We were doing that so C clients could be more efficient (and easier/safer to write). - Brad On Thu, 17 Aug 2006, Steven Grimm wrote: What did you have to leave out due to conflicts? -Steve Brad Fitzpatrick wrote: This is a merge of facebook's changes into the main version, with some of the new features that were conflicting removed, others partially still there, etc, etc. WARNING: we're not using it in production. I can't vouch for its quality. Facebook can vouch for the version they used, but maybe I messed it up merging it in. HERE BE DRAGONS. Be scared. That said, http://www.danga.com/memcached/dist/experimental/memcached-1.2.0-rc1.tar.gz Notice the /experimental/ in the URL? PACKAGERS: Don't package this yet. It's a dev release. Everybody else: bang on it. Let's find any issues and get a release out. It's been too long, and I want to finish vbuckets, but want a stable tree to hack on. Also wanted: benchmarks between this and last stable release. Feedback requested.... - Brad