You may have noticed (ha!) that K5's been slowing to a crawl around peak times, and that peak times have metastasized to encompass most of the workday EST. The basic problem is that the database machine (a dual 700Mhz Pentium III with 1Gb of memory) doesn't have enough memory space to sort a few of the key indexes. In particular the time index from stories and just about any index from comments.
In addition to this, we're one box short still, as one of the Scoop servers suffered a hard drive failure a while ago and isn't back online yet. The plan is to put a larger hard drive in the former Scoop box (a dual 1 Ghz P3 with 3Gb of memory) and make that the database server, which should be able to speed up the DB a lot. Then the former DB machine will become the third Scoop server, joining the other two more powerful boxes. This should help a lot. Rest assured that I am bugging Voxel to get this finished as soon as possible. Also note that blaming Voxel for this is rather out of line, since they are (still) providing all of the bandwidth, hardware, and support for free.
Meanwhile, I tweaked some of the database variables this morning to allocate more memory for index and sorting buffers. I'm kind of crossing my fingers and praying that it doesn't blow up, since that machine doesn't really have more memory to allocate, but it seems to be helping so far.
No, search isn't fixed yet. K5 is a few updates behind the current code already, and with the hardware changes that are definitely upcoming, I thought it made sense to move the boxes around and update Scoop all at the same time, so as to minimize total downtime. Search won't be fixed until we're up to date with the current codebase, because I'm definitely not going to hack in some K5-specific fix that isn't in Scoop CVS.
I think most of the other things people have wanted are in the current Scoop CVS, like the fixes to the edit queue and a bunch of bugfixes to autoformat posting mode.
So the sane order for updates is fix hardware, update Scoop, squash the inevitable bugs and misfeatures that always show up in Scoop updates, then patch up the search. I should have at least a first try at a search fix by the time K5 is ready for it.
The CMF bylaws discussion is proceeding, though somewhat slowly since everyone on the Board has a lot of other things to do as well. The big questions we're dealing with right now:
I think that what we come up with in answer to these issues will become the first public-review draft of the bylaws, so hopefully we'll have a version that's appropriate for posting here in a few weeks, at most.
- Membership dues: I'm proposing these should be voluntary, not mandatory, with language to that effect in the bylaws
- Directors: The first draft had, I think, a poor design for the Board. It proposed nine directors, of which four would be elected and five appointed, with the appointed Directors in turn appointing their successors. I think this puts far too much power (like, all of it) in the hands of the non-elected Directors. My proposal is that five Directors be elected, and four appointed, and that the appointed Directors should be appointed by a vote of the entire Board.
- Miscellaneous things: Some details about requiring the annual report to be posted online as well as filed at the organization's offices, issues about how and where to hold membership meetings, questions about what constitutes a quorum for both membership meetings and Board meetings, and a couple of basic editorial issues.
Rusty Doesn't Care Anymore
This theory has been making the rounds, much to my consternation. I don't know where it comes from, but it ain't true. If anyone would like to fill me in on what I did to make them think that, I will attempt to stop doing whatever it is. Or, um, start doing whatever it isn't. If that makes any sense.
Please feel free to complain below, if you need to get it out of your system.