We're migrating the data in Onaswarm to a high performance network attached disk ... we expect we'll be back by 8 AM EDT.

We did a significant update to Onaswarm this morning:

• Everywhere where we used to refer to "feeds" we now refer to "social networks". We believe this may be a little more clear
• We've simplified the "friend" relationship (which is really more like the Twitter "follows" idea, so we may be changing that more soon) so you don't choose to "aggregate into my account". Rather, you'll see to the right links for "Calendar", "Lifeblog" and "Lifestream"; next to those is a "+ friends" links, which you can click on to show what your friends are up to
• We've made a number of database improvements that should make things a little faster

We updated Onaswarm again this morning:

• country flags (based on Swarm membership) are displayed on the main page
• we can now autodiscover Upcoming feeds -- this was somewhat difficult before because they don't have an easy (i.e. non-API) way of converting account names into their numberic IDs.
• Tumblr is now autodiscovered
• A bug in parsing WordPress feeds has been fixed

We made a significant update to Onaswarm this morning. The update contains almost no new features, just updates to usabilty:

• the Administration page has been substantially simplified
• the "button exploding" effect is gone
• almost all button actions now grey the screen and display a waiting symbol
• all visual effects are programmed to happen faster (i.e. 100-200 ms) to make the site snappier
• a number of IE and Opera related bugs were fixed

As per usual, your feedback is always appreciated. Do not hesitate to contact us if you have suggestions,  bug reports or feedback.

We updated Onaswarm this morning.

• we added Digg as a recognized feed provider
• we added Digg and Del.icio.us (d'oh) to the list of feed providers we look up
• internally, we added feed specific filtering to remove a lot of junk Digg adds to feeds
• we improved the process for adding friends (a surprising amount of work)
• we improved the process for adding feed
• we improved the process for joining swarms
• we added the ability to change post dates and save posts as drafts
• we limit the number of similar entries shown in the standard summary view (I'm looking at you, Last.fm)
• many tiny tweaks and bug fixes

We've had a busy Saturday morning, including a short crash due to load. We're having a look at our kernel parameters to see if we can identify what ran out so we can just slow down rather than fail!

We are upgrading to V11 this weekend. This may involve downtime, especially Sunday morning though we'll try to minimize disruptions. At times may be slow since we'll be effectively running two parallel BlogMatrix.coms for a while.

Update: we only got half way through the update and had some issues. We'll be picking up the second half ASAP.

We successfully upgraded the software to it's latest version this morning. This is mainly UI tweaks and performance enhancements -- if there's any issues, please contact us ASAP.

It's been a while since we've reported in. Things have been running relatively smooth since the last release. We've identified a number of improvements, mainly related to the database, and this seems to have solved 90% of our problems. We're looking to hit the last 10% with:

• reducing the size of log files (which are killing cache efficiency)
• having certain applications monitor system state and pausing during busy periods
• running "headless" (without front end apaches, which are causing the page cycle problem) 

We've just completed a major system upgrade that (hopefully) will fix all the problems we've seen for the last 8 days.

Note: some accounts may be being served as "[.cheetah]" -- this will be fixed shortly.

Update: it failed. The system just grinds to a halt; we're starting to seriously suspect we have hardware problems.

Our attempt to upgrade this morning partially failed: we got all our software updates in, but the big "reorganization" of how our Apache servers work didn't take for some inexplicable reason. So we're going to try again tomorrow morning. Unfortunately, this means that getting the "wrong page" is still going to occur (though much less often than you were seeing 4 days ago).

As you've probably noticed, we're still having problems. The issue is with straight-out-of-the-box Apache 2.0.XX which acts as a front-end to our app servers. Our work around right now is to restart these frequently, but this isn't nearly a perfect solution. Our long term solution is to go "headless", meaning no front end servers. We've got this 98% working but the devil's in the details with this one. This long term solution may be in first thing tomorrow morning, with a little luck. If not, we're upgrading to the latest latest version and seeing if that will work.

For future Google reference, these keywords may help: apache, mod_rewrite, mod_proxy, corrupt, mangled URLs, regex.

We had another few issues yesterday afternoon, but we think we've got the problem licked. Here's what we've done:

• our front end "primary" apache servers seem to be getting into a corrupted state. Why this is happening is a mystery to us, but mod_rewrite seems to be combining bits of different queries together, just giving the "cycling through pages" issue. We've "solved" this by frequently cycling these servers, which isn't a great solution but it's good enough for showbiz. We're tracking this deeper.
• we've improved our core query (essentially, find an entry with a set of tags, some of which are very common) by 50x. Even out-of-cache page loads now are sub-second, so we're very excited by this to say the least.
• we had a problem with Internet Explorer browsers and posting -- we've fixed this, and if you're still seeing a problem, trying reloading your page to get the latest JavaScript includes
• We've made a number of UTF-8/Unicode fixes, so you can now view (and potentially edit) pages that look like this.

We had a major spot of downtime yesterday, from approximately 17:00 - 20:00 EST. The issue was caused by a series of bugs that compounded upon themselves, eventually grinding our serves to a halt. We've fixed these issues and things seem to be running smoothly, if a little slow today. The issue right now is certain parts of our caching code -- which is one of the key ingredients for speed -- seems not to be always triggering, causing a certain query to be executed over and over again when it doesn't need to be.  We'll try to fix this today in development and put the fix into production early AM tomorrow.

Sorry for all the problems.

We were playing with our Apache config files this morning. as certain pages weren't taking the new "grey" look we've been doing. Sorry for any disruption.

Via Chris Pirillo, we learn of Monastic "the free website monitoring service that doesn't suck". They check your site every couple of minutes and if it isn't available, they send you an e-mail and notify and RSS feed.

They should add SMS-message and charge as a premium service. Why? Well, (a) your cell phone is probably where you want to get the message and (b) when people/companies make money, the service gets longevity!

Our webserver was shutdown for several hours this morning – stuck rebooting – for reasons we are trying to find out

We were down (and up) several times in the last 12 hours. It turns out the problem was a burned-out CPU fan on our “www” server which was quickly heated the processor to “boil”. We’ve replaced it and everything should be running nice and smoothly again.

If you’ve been having problems downloading in the Toronto area, this may be the problem (þ Boing Boing). We have no solution at this point, except to suggest only downloading one item at a time.