Server down for maintenance

[wirthi]

Hi,

I sometimes (rarely, but it happens) get this message when I start the client: "Server down for maintenance". Then the client waits for user-input and doesn't do any work. When I restart the client, everything works just fine.

So, why does it not just start to fold? At least since you accept "old" buffers for some time after a protein change, the client schould continue its work, even when it assumes (I guess it does) that a protein change is in progress.

Greets,
Wirthi

[Brian the Fist]

This is likely happening when the server is overloaded with work. If the client for some reason tries to verify the status of the server but doesn't receive a reply (before timing out) it will conclude that the server is down for maintenance and exit.
It is designed so that you cannot START folding while the server is down/very busy, but if you are already folding and just at a checkpoint (every 5000 or so) it will keep going of course.

Thus this behaviour is intentional.

[wirthi]

First thanks for your quick reply!

I guessed that this is intentional behaviour, I just don't understand why. I guess by not letting my client start you want to reduce the amount of work the server has to handle (right?) Why do you do that when I start the client? I mean, if the server is not able to handle more *uploads* when I try to upload, that's ok (the client buffers the work, and after another 2000 folds it tries again).

Another argument would be that when I start my client (and there is no new version to be downloaded and I won't upload any old buffers) the first time my client stresses the server is after 2000 folds (that's more than 3 hours on my computer). So, if the server is currently not able to accept new uploads, that has nothing to do with my client, since it will upload in the "far future" for the first time!

Just another one: everytime this happens I just restart my client (say, 10 seconds after the first try) and then it works perfectly. I really don't feel like working as babysitter for the client, just because the server is busy working for some seconds during my startup.

[Paratima]

On NT / W2K / XP systems, running as a service, the client starts just fine even when it can't contact the mothership. This is even true when restarting the service "manually", via Settings - Control Panel - Services - Start Service. Seems as though you ought to be able to restart it manually.

[wirthi]

Originally posted by Paratima
On NT / W2K / XP systems, running as a service, the client starts just fine even when it can't contact the mothership. This is even true when restarting the service "manually", via Settings - Control Panel - Services - Start Service. Seems as though you ought to be able to restart it manually.

I *am* most of the time able to start, even when there is no connection to the server. Only very seldomly (it happend at max 5 times since I started folding in april) the "server is down for maintencance". Since the client keeps running (but does not produce any work) it is really annoying ....

(of course I didn't lose very much, but I lost something - and that could be prevented in my eyes by just starting to work, even when the server is down)

[Brian the Fist]

I have learned that no matter what I do, some people will find something to while and b*tch about (nothing personal here :D ). Thus when people make suggestions, I try to put myself in their position and ask myself, would it make my life easier if I made that change? If the answer is yes, I change it. If not, I leave it alone. I generally do what will be best for our server and network and database setup. For the most part I have received excellent comments and suggestions so do keep them coming, but don't be offended if I ignore some of them. All are considered.

[wirthi]

I have learned that no matter what I do, some people will find something to while and b*tch about (nothing personal here :D ). Thus when people make suggestions, I try to put myself in their position and ask myself, would it make my life easier if I made that change? If the answer is yes, I change it. If not, I leave it alone.[

Hey, I take that personally ;) I am not b*tching about, I just found something that annoys me. That doesn't mean anybody else must have the same troubles. For ME it is bad to loose structures in the stats, just because the client doesn't like to work if the server is busy (I still don't see any direct connection ...).


I generally do what will be best for our server and network and database setup. For the most part I have received excellent comments and suggestions so do keep them coming, but don't be offended if I ignore some of them. All are considered.

I really appreciate your work Howard, the personal service you offer (especially here in the forum) is great, compared to other projects. And if you descide the suggestion I have made is not worth being implemented, that's ok for me, you are the one who is responsible for the stability of the project.

But I still keep the right to tell you if I find something annoying - and losing structs is the most annoying thing that can happen while helping a DC project (apart from getting the HDD formated by a buggy client).

[Brian the Fist]

Since the problem only occurs when you are starting the client, you are sitting in front of the computer. Thus I hardly see how it is an inconvenience or how you are losing structures if you simply need to try starting it 2-3 times before it succeeds. If the server is truly bogged/unavailable, we do not want you to be able to start the client (call it a control issue) so I think it is a small inconvenience at worst, and as you said, relatively rare (and will be even less rare soon when we have a more powerful back end we hope). Anyways I'm not making any changes of this type until the new backend is in place and we see how it holds up.

[Mikus]

Here's hoping that the back end gets improved soon. Tuesdays the server is extra busy because it is downloading the new protein to the clients. Well, I just uploaded a fileset to the server, but (presumably) because it was busy it never acknowledged the fileset -- my client timed out and resumed folding the previous protein. I wish that the time-out in the client was long enough so that even Tuesday backlogss did not cause the client to give up.