As long as it takes. This is not a problem.
When the experiment is over we’ll fold his posts in under his old name.
Perfect. Thanks.
You do:
(bolding mine)
Your database is hosted on a cluster of five servers, so if one of them is doing something weird that could explain the randomness of the problem.
This would be a software sync (i.e. running a program that keeps the computer’s clock synced with a standard time, e.g. NIST), not anything to do with the physical hardware. So as long as he can remotely log into each server, he can just check the system time and make sure they’re the same. (I’d be very very surprised if they aren’t, that’s pretty standard, but different computers disagreeing about what time it is could explain Siam Sam’s problem.) This also might be something he just wants to call to DesertNet’s attention, since the database cluster is a shared resource: maybe they can do some standard diagnostics to make sure each server in the cluster is working properly and able to handle mySQL queries?
It’s a very good question. The root cause may be a random but unlikely occurrence, e.g. the problem only happens when your last logout was processed by database server #3 and your login is processed by database server #5. Some users may be doing something to trigger this event more often. For instance, I personally suspect the fact that Siam Sam is constantly logging out and logging back in has something to do with it – most users just leave themselves logged in on whatever computer(s) they use. If the problem occurs most often during login, only a small fraction of posters will even have a chance to see it. Add in the fact that posters who don’t use the New Posts feature could be hit by it and don’t notice or care, and your pool of people who are hit by this shrinks down to a handful.
It is difficult to know the why before we know the how. Tell me how the last visited is being updated when it shouldn’t, and I’ll have much better guesses as to why.
But for a couple of WAGs…
If we have a couple of servers, then we have some method of load-balancing. Typically there’s an actual load balancer appliance that determines who goes to what server. That determination is algorithmic, with the inbound IP address often being a prominent part of the calculation.
If so, you can have a situation where IP address A goes to server A and IP address B goes to server B, assuming everything’s up and working fine. But let’s say there’s some intermittent brief problem that interrupts the connection between the load balancer and server B. Maybe it’s brief enough that nobody gets alarm klaxons on their desk or whatever, but in that brief time, IP B would be sent to server A–so if there’s any sync delay people on IP B might get zapped, while people on IP A never notice anything because there’s no problem there.
The sync itself could be something too. I have no idea how any syncing might be done, but it’s probably not out of the question to speculate that you, or Jerry, or many other admins have “earlier” user accounts and might sync “faster.” Or maybe the people who designed the sync already tried to take that into account and it goes every fifth account or something, so if you happen to have an account in the last batch, you’re more likely to be out of sync.
Again, these are WAGs, but there are all sorts of ways that only subsets of the population can get hit with something like this.
Thanks for all the explanations.
Waiting to hear from Jerry.
As a concrete example - one of the symptoms we had here a while ago was I would open an SDMB tab and have my last visited time set to right now. But if I refreshed the tab, it would go to an appropriate older timestamp. This is probably because I was bounced to a different server on the refresh.
Sadly, that doesn’t seem to work with the more common zap.
Also, in light of Giraffe’s hypothesis that Siam Sam’s issues are exasperated by logging out and back in, I’ll re-iterate that I never log out, and I get zapped on a semi-regular basis.
Doesn’t mean the hypothesis is wrong, mind you. I’ve seen lots of odd interactions that only made sense once we found the root causes. Just keeping the data up to date.
…and, just zapped. I was offline Wednesday, popped on today around 8pm, and caught up on my threads. Went around the house and did a few chores. Came back to see if there was anything new. When I got to the forum list, it said my last activity was 9:23. I opened my regular forum tabs, and got…ATMB 9:23, Cafe Society 10:09, GQ Internal Server Error (reloaded page and was marked at 10:10), IMHO 9:23, and MPSIMS 9:23. In that order.
Since I was already zapped, I hit refresh on my Cafe Society page a few times. It came up at 10:10, 10:10, and then paused for about 30 seconds in the third refresh…but came to a 10:10 timestamp eventually.
OK, I’ve done a bit of hunting around in the code and the AdminCP. Before I expound on how the lastvisit is stored / accessed, I have a key question for Tuba / Jerry:
What is your AdminCP / vBulletin Options / General Settings / Thread/Forum Read Marking Type option set to?
There are three options:
Inactivity/Cookie Based
Database (no automatic forum marking)
Database (automatic forum marking)
Followed by another relevant field: Database Read Marking Limit, the time in days to store the thread and forum read times.
No point is speculating about anything else until we know the answer to this question. If it’s cookie based, that puts it on the user / browser. If it’s database based, that puts it on the server.
I do believe it’s Inactivity (cookie based).
(If memory serves.)
Can you check?
I’m curious how cookie-based even works, every thread I had looked for indicated it was server-side.
Looking at the cookies for this site, I found two that looked related, bblastactivity and bblastvisited. bblastactivity had a value of 0, and that value stayed at 0 throughout a session of browsing the boards. bblastvisited had a number string in it, but that didn’t change either. After my first session, I waited for about an hour and then came back–the boards correctly showed my Last visited as updated to the end of my last session, but the number in bblastvisited was still the same.
During this session, I deleted the bblastactivity cookie, it didn’t change anything. I then deleted bblastvisited, still no zap or any other change that I can see. Neither have come back yet.
I’m not entirely clear what ‘zapped’ refers to. IF it’s about logging in and having nothing marked unread because your last visit is basically ‘Right Now’, then that happens to me with such frequency that I’m surprised it works for anyone else.
So I guess that’s what people mean in other threads around here when they say they only surf new posts, which always seemed like a ridiculous redundancy to me since obviously that’s how a person surfs a message board. But perhaps they are in fact referring to a feature that wouldn’t serve someone at all that can’t reliably count on the board’s software to provide them an accurate account of any new posts since one’s last visit. Overly long run-on sentence aside, I have an issue similar to Siam Sam’s, I think.
I only use Firefox, I only log in on this PC (barring my phone), and I have recently moved to a new zip-code, with a completely different internet provider, which in no way eliminated the problem. If that data helps at all.
Yes, that’s what it is. Some of us have reported the problem. Some people never seem to experience it and others, we sufferers assume, do experience but don’t report it.
Saturday morning, and I’m logged onto this account in Firefox. Not zapped. Real account is on in IE and not zapped either.
I’ve been reluctant to use just this one sock account because of the threads I’ve been following and the desire not to confuse other posters. I thought maybe if I used both accounts in different browsers, that might give some useful data such as the fact that so far in this experiment, I’ve been zapped only in Firefox whether it’s my real or sock account.
But after I log off of the real account today, I’ll leave it alone for a while and just use this one, although I might still mix up different browsers for it.
Okay, I just logged out of me real account and will use only this sock account for a while.
Do you think I should stick with just one browser? Or mix it up a bit – Firefox, IE, Chrome?
Popping in to see how it’s working. Not zapped this time either. Using Firefox.
Heading out and will check back in tomorrow.
Unless someone feels strongly otherwise, I’d stick with one browser. We should change one thing at a time. Do everything as if you were using the main account, but use 222 instead. If you still get zapped, then will assume it isn’t as likely to be associated with certain accounts. If you don’t, then it seems like an account-based issue…at which point I think we should duplicate the experiment with another sufferer to see if a sock resolves their issues as well.
But for now, get some stable data and we’ll see what it means.
Okay, I’ll stick with my usual browser at home, which is Firefox. But again, since I’ve had this sock account, I’ve been zapped once on it and once on my real account, both on Firefox, although in the past I’ve been zapped on IE too.
just spent 5 mins trying to log on and getting a message 503 service unavailable and telling me there is no server???
Every Sunday at 2am CST in the US, the Board does something like back itself up. It’s down for about a half hour at that time. That’s probably what you encountered, as it would be about the right time frame.