Is this the end?

I'm concerned about the future of this website.. I play here for more than 3 years, and I played very different games (started with hex). Never in this time, ever, I had some few games with -135h, championships do not start even though there are some dozen people waiting, empathy game is late for more then 100h... What is this trying to say? Is this the end of littlegolem?

no

LG was off for six months a couple of years ago and came back. So I think positive.

But hopefully the current problem will not last this long. ;)

It was never off for six months. We did lose about six weeks of data once, though. Other than that, we've had daemon outages like this ranging from a few hours to many days.

I'm worried though. The amount of negative time accumulated on this server is enough to provoke an anti big bang in reverse, thus ending the universe. I hope there are no leaks.

the major data loss was in the summer of 2006. There has been nothing too serious since then, except the daemon outages, which happen too often. (Programmers-is this possible to fix? I would think it should be easy, but I am not a programmer. Perhaps set up the system to reboot the daemon automatically every 24 hours?)

This site could be self sufficient if this problem was fixed permanently, but as it is, it is dependent on Richard rebooting the system. Richard has 70 games in progress, and some have negative time. Hopefully, he is just away on vacation and will return soon!

Ok, only six weeks. Maybe it felt like six months...

Ak je demon, ktory to ma byt spusteny po celu dobu abends a pricina Abend je potrebne analyzovat pred opatovnym to bude na nieco dobre, potom nahradzat demon s cronu a naplanovane ulohy, ktore bezia 2 az 3 krat za hodinu , konci, akonahle je to hotovo, a znizit ucinok niektorych uloh, ze je to abending na, randomizes poradie, v ktorom sa riesi rozne ulohy, odkazany prinajmensom umozni vela miesta pokracovat, aj ked abending ulohu este neurobil, kym nebude mozne analyzovat rucne.

a od 2010-05-14:

Myslim, ze sposob, ako to funguje teraz, napriklad softver zapisuje zaznamy do databazy ako niektore hracov zaregistrovat mesacny turnaj v urcitej hre. Potom sa demon pozrie na zaznamy, kazdu hodinu alebo tak, a ked je ich tam 5 alebo viac zaznamov pre tuto hru, zacne turnaj pre 5 hracov, ktori sa zaregistrovali, ale este neboli zaradene turnaja.

Demon bez moznosti by bolo, aby softver, ktory vlozi novy caka hracov zaznam skontrolovat, ci tam su teraz caka 5 hracov, a ak tam su, zacat novy turnaj okamzite bez cakania na niektory daemon. [To by bolo ako kod, ktory vydava spravy, ked zapas skonci, alebo skonci turnaj. To necaka na ziadne daemon. Tieto spravy su este vydane, aj ked demon nefunguje.]

Ano, musite serializacie pre tento pristup (viac ako jeden-vlaknovy daemon potreby), ale vzdy som bol schopny prist s dizajnom, ktory je v databaze serializacie zariadeni dostacujuca (napr. vsetky spory mozu byt vyriesene pomocou databazy vlastnost, ze iba jeden vlozenie noveho zaznamu s rovnakym primarnym hodnot index pola moze uspiet.).

Je pravda, ze v pripade, ze problem je chyba, s ktorymi sa turnaja start kod by daemon bez kodu maju rovnaky problem. Ale myslim, ze to je celkom nepravdepodobne, ze vsetky demon cinnosti, ktore by mohli byt daemon bez bude sa stretavat s chybami v rovnaku dobu, zatial co ked demon zastavi, vsetko sa robi sa zastavi, a to nielen cinnosti narazi na chybu.

JSP, ktory zobrazuje vasa hra empatie odpovedi zacat novu hru, ak stare hry lehota uplynula. Ak sa nikto pri pohlade na GoE stranke, potom nova hra nezacne "vcas", ale tento fakt by nepozorovatelne. (No, takmer nepozorovatelne. Dalo by sa tiez dal spusta v podrobnej stranke s vysledkami, a v poslednych hier stranke.)

Aktualne demon moze byt, ze zacat kazdy den o polnoci, alebo kazdych 10 minut, cron naplanovane ulohy, alebo inak. Ak sa kontroluje plnenie uloh, ktore mozu vykonavat v niektorej deterministicke poradi a prva uloha sa pokusa je ten, v ktorom doslo k chybe, ktora sposobi, ze ukoncit alebo cakat vecne, alebo nekonecny cyklus, alebo co by sme vidiet rovnake priznaky ako daemon, ktory sa spusti iba rucne.

@Ray, yes, the daemon could be eliminated.

To quote myself from 2009-09-21:

If a daemon that's supposed to be running all the time abends, and the cause of the abend has to be analyzed before restarting it will do any good, then replacing the daemon with a cron job or scheduled task that runs 2 or 3 times an hour, ends as soon as it's done, and, to reduce the effect of some task that it's abending on, randomizes the order in which it tackles its various tasks, would at least allow a lot of the site to proceed, albeit with the abending task still left undone until it can be manually analyzed.

and from 2010-05-14:

I guess the way it works now is, for example, the software writes records to some database as players sign up for a monthly tournament in some particular game. Then the daemon looks at the records every hour or so, and while there are 5 or more records for that game, it starts a tournament for 5 players who have signed up but not yet been assigned a tournament.

A daemon-free alternative would be to make the software that inserts a new waiting-player record check to see if there are now 5 players waiting, and if there are, start a new tournament immediately without waiting for any daemon. [This would be like the code that issues messages when a game ends or a tournament ends. It doesn't wait for any daemon. Those messages are still issued even when the daemon isn't working.]

Yes, you need serialization for this approach (more than a single-threaded daemon needs), but I've always been able to come up with a design that makes the database's serialization facilities suffice (e.g. all conflicts can be resolved by using the database's property that only one insert of a new record with the same primary index field values can succeed.).

Granted, if the problem is an error encountered by the tournament-start code, the daemon-free code would have the same problem. But I think it's fairly unlikely that all the daemon's tasks that could be made daemon-free would be encountering errors at the same time, while when the daemon stops, everything it does comes to a halt, not just the task encountering an error.

The jsp that displays your Game of Empathy answers could start the new game if the old game's time limit has passed. If no one's looking at the GoE page then the new game wouldn't start "on time", but that fact would be unobservable. (Well, almost unobservable. You could also put triggers in the detailed-results page and the past-games page.)

The current daemon may well be being started every midnight or every 10 minutes, by cron or Scheduled Tasks or otherwise. If it checks the tasks it can perform in some deterministic order, and the first task it attempts is the one in which it encounters the error that causes it to quit or wait forever or loop forever or whatever, we'd see the same symptoms as a daemon which is started manually only.

Or maybe Richard is just preparing backgammon?! I hope

It IS vacation time. Unless the'res still a problem come August, I'm not worried at all... :-)

I used to worry about those things, but I do not anymore. During THE BIG CRASH, Richard's laptop broke while he was on a vacation. Richard has proven to be reliable and dedicated enough to ensure the future of LG.

I agree with Marius. August is a good time to start worrying. Happy vacation all:)

in each (possible) end there is the seed for a wonderful

new start.

Ingo.

Although in the long term I won't worry too, in the short span it might reduce a lot LG's fun with midterm reflexes (due to reducing the number of active players). And, may God forbid, a new crash would potentialize those effects. When the BIG CRASH occurred I left the site for several months, ensadned that this went void.

*ensadned = desgostoso (pt-br)

@DoP - for me, after the poll, having backgammon here in LG is a near obligation for Richard. But I wouldn't mind at all if when he does so he also include a game he's found of or want to promote (say X-Chess, for example). This was the way I came to know such great games as Breakthrough and OSKI (in which I totally suck but love it nevertheless, I've even played a few pen and paper Portuguese Oski games with my chess students). I wouldn't mind having such a game even if its still in a preliminary stage of development and the rules will have to change in the future (as it happened to Wyps).

Well once games in progress are finished, it will be the end till the time is fixed. Only way i have been able to start a new game is by sending invitations for unrated play.

I'm 3th at last GOE-game, and 2nd over the last 7 games. The longer we're stuck, the longer my succes is shown!

(Just kidding, off course I hope it will be fixed yesterday).

Richard has made a move at 15.29 servertime so he has not gone!!

You are positive on this? Than he might be playing from his mobile and not able to restart the daemon, but still alive...

Great

http://www.littlegolem.net/jsp/game/game.jsp?gid=1329378

Maybe Richard will make this move next week, and a leak in the negative time storage room made the move appear now?

i think DukeOfPrunes is right. i've noticed that often when this happens new games are introduced.

The world has started turning again. Has anything changed?

The site went into maintenance briefly (I then left home to the supermarket) and then came back. So far I haven't noticed any change but the daemon restart.

all the minus times are gone, and GoE is started. Also, Championships started.

So, everything is back to normal, it seems.

I confirm what Dvd Avins stated above. Golem was never down for six months, nor six weeks, nor even six days. At the most, I think maybe Little Golem might have been down for a day or two. But yes, approximately six weeks worth of DATA was lost a couple of years ago. (Which still bothers me... I was tearing everyone up in Empathy during that time. I was on a roll, and I don't know why. Lots and lots of my wins were lost.)

I think we were actually down for a couple of weeks, but I'm not sure of the details.

The longest outage I recall was the one after which s cript started becoming s-cript. Vulnerabilities to cross-site-scripting (and maybe SQL injection?) got fixed while the site was down.

We're working again!