24 hour rally

[gd_barnes]

Originally Posted by Anonymous

Or, maybe someone else could volunteer to run one of the LLRnet servers. I would gladly run a few servers, except that I don't have any machines that are consistently on 24/7.

Anon, I need to learn to do this. So I'll dig in and see what I can do about having you train me on setting up a server next week. I'll get with you. I still have the long detailed PM that you sent me way back as a starting point.

I have 3 cores on 24x7 but I have been known to have Road-Runner go down on occassion. It just happened last night for a few hours but that was the first time in a few weeks. I estimate that I lose the connection perhaps on average of twice/month. The only problem is, I'm not quite as flexible as Carlos situation-wise. If it goes down at 10 AM, then we're out of luck until I get home from work around 6:30 PM.

What we could consider is having me run 1-2 severs but not for rallies. We'd still do the rallies on Carlos's server but have him running no more than 2 servers while doing so to avoid memory issues.


Gary

[gd_barnes]

Originally Posted by Anonymous

Okay, that sounds like a good idea. That way we can still do the rally on port 100!

BTW, if you want to queue more than 100 k/n pairs at a time, all you have to do is make a small modification to the llrnet.lua file. I've attached a replacement llrnet.lua (just remove .pdf from the end of the name and replace the one in your LLRnet folder with it) that will let you queue up to 1000 k/n pairs at a time.

EXCUSE ME??!! Port 100 has been down for hours. It has a memory leak.

I assume you're only joking. I have all home machines set to start on port 300 and I'm ready for bed shortly.


Gary

[mdettweiler]

Originally Posted by gd_barnes

EXCUSE ME??!! Port 100 has been down for hours. It has a memory leak.

I assume you're only joking. I have all home machines set to start on port 300 and I'm ready for bed shortly.


Gary

Actually, I wasn't joking...I was thinking, since the port 100 server seems to be coming back on a bunch now...oh well, I guess you're right, we still should do it on port 300. (I'm really tired, it's 3:00 AM here.)

Edit: Okay, I've got both my cores set to queue 400 k/n pairs at a time, on port 300 of course.

[mdettweiler]

Originally Posted by gd_barnes

Anon, I need to learn to do this. So I'll dig in and see what I can do about having you train me on setting up a server next week. I'll get with you. I still have the long detailed PM that you sent me way back as a starting point.

I have 3 cores on 24x7 but I have been known to have Road-Runner go down on occassion. It just happened last night for a few hours but that was the first time in a few weeks. I estimate that I lose the connection perhaps on average of twice/month. The only problem is, I'm not quite as flexible as Carlos situation-wise. If it goes down at 10 AM, then we're out of luck until I get home from work around 6:30 PM.

What we could consider is having me run 1-2 severs but not for rallies. We'd still do the rallies on Carlos's server but have him running no more than 2 servers while doing so to avoid memory issues.


Gary

Okay, sounds good. You are now offically enrolled Professor Anonymous's LLRnet server administration classes via mersenneforum.org private message!

[gd_barnes]

Originally Posted by Anonymous

Actually, I wasn't joking...I was thinking, since the port 100 server seems to be coming back on a bunch now...oh well, I guess you're right, we still should do it on port 300. (I'm really tired, it's 3:00 AM here.)

Edit: Okay, I've got both my cores set to queue 400 k/n pairs at a time, on port 300 of course.

Carlos should make the call here. I wasn't trying to influence him earlier when I said port 300.

Whichever port Carlos views as more stable is the one we should do as long as he can get all the Free-DC folks notified in time.


Gary

[gd_barnes]

Carlos,

Port 300 is back down. I give up and am going to bed. I should be back on around 11 AM (5 PM GMT).

My son's machine will sit dormant waiting on port 300 until it comes back up and I'll be able to tell what times it was processing when I get up.


Gary

[em99010pepe]

All servers are up and running...

[mdettweiler]

Originally Posted by gd_barnes

Carlos should make the call here. I wasn't trying to influence him earlier when I said port 300.

Whichever port Carlos views as more stable is the one we should do as long as he can get all the Free-DC folks notified in time.


Gary

Well, it might be a little too late for me to do port 100 now...now that I've got 400 k/n pairs queued up for each core.

If Carlos does decide to do port 100 for the rally after all, I can still switch over tomorrow morning (not going to do anything more tonight, gotta go to bed!), and then finish up my big port 300 queues after the rally.

[mdettweiler]

Originally Posted by em99010pepe

All servers are up and running...

Okay, good. I'm going to leave my two cores on port 300 tonight, since they've both got tons of work queued--but I'll move them to port 100 tomorrow morning. (In my time zone, the rally starts at 2:00 PM.)

[em99010pepe]

I think we can still make the rally on port 100, now everything seems to be normal.

[em99010pepe]

Originally Posted by kar_bon

yep, so to me: 300 ok, 100 down!
hope carlos won't sleep tooooooo long

Didn't I give you my cellular contact? Next time send me a text message....lol

From now on I'll add small junks of work to the servers, of course this memory leak will vanish as we climb on k/n pairs, they start to take more time to process and so less connections between clients/server.