Ups and Downs - The Server Thread

[loftar]

I'm back, and I've fixed the bug that caused the crashes while I were gone. The reason I'm able to salvage data is because I can run the server in a debugger and (usually) recover a state reasonable for saving manually, which is also the reason why the server cannot (or, at least, should not) be restarted automatically.

[Harleyvitality]

Huzzah! So when is the next Dev session?

[Voideka]

I'm back, and I've fixed the bug that caused the crashes while I were gone. The reason I'm able to salvage data is because I can run the server in a debugger and (usually) recover a state reasonable for saving manually, which is also the reason why the server cannot (or, at least, should not) be restarted automatically.

Why do you need to save manually? I'm curious as to how all this works.

[Pacho]

Because if it saves corrupted data automatically and we play with that, we get a whole mess of problems and bigger rollbacks =|

[loftar]

Why do you need to save manually? I'm curious as to how all this works.

I donät need to save manually. I'm just talking of the case of an actual crash, when I can usually do some manual work to get the game data saved before I restart the server.

[Voideka]

I donät need to save manually. I'm just talking of the case of an actual crash, when I can usually do some manual work to get the game data saved before I restart the server.

That's what I was talking about, yes. Is it to try to prevent a big rollback?

[PhaedrothSP]

Did someone trip on a cable?

[StoreClerk]

What are the specs on the server?

Also, can we get a graph of crashes over a monthly period?

lolololol

[loftar]

Did someone trip on a cable?

No, but as mentioned in the Bugs forum, I'm having some trouble with the new account server that I just put into operation. I'm hoping to have it fixed very soon, but the bug keeps eluding me.

What are the specs on the server?

The current server is an AthlonXP@1.8GHz with 1 GB of RAM. (But it would need another GB to run Apache properly)

Also, can we get a graph of crashes over a monthly period?

Given how I wouldn't trouble myself to maintain such a graph, no. :)

[skrylar]

Edit: Artist's impression of haven server.

1) I still say you should take a look in to tokyo cabinet, it runs on the good server OS' and it has an order of magnitude faster response times than what you previously mentioned using (GDBM) for probably the least amount of effort put in.

2) Suggestion: Set up a separate, obviously smaller database which would server as the party time server which gets activated from a pristine snapshot every time the server crashes. There is no point to this, it would just be a place full of arbitrary things the size of only a few grids and would let people run around trying to capture chickens or newbie swarm dragons. Again, no point at all other than it would be amusing during downtime and possibly being able to capture more error logs if people try to use the time to perfect their server crashing techniques. A test server if you will.