poclbm-mod crashes spontaneously.

[kinghajj]

I've been mining for a couple weeks now, but have run into a problem. Occasionally, my miner program (poclbm-mod) will "crash": it stops sending work, yet for some reason still uses GPU time! You can imagine this really sucks if it happens when I'm sleeping... so many wasted hours. Anyways, has anyone else had this or similar problems? I'm running it in Linux with a 5870, if that matters.

[1714958604]

1714958604

[1714958604]

1714958604

[1714958604]

1714958604

[[Tycho]]

I've been mining for a couple weeks now, but have run into a problem. Occasionally, my miner program (poclbm-mod) will "crash": it stops sending work, yet for some reason still uses GPU time! You can imagine this really sucks if it happens when I'm sleeping... so many wasted hours. Anyways, has anyone else had this or similar problems? I'm running it in Linux with a 5870, if that matters.

May be it's a GPU problem ? Insufficient cooling or faulty card.
Try another mining program and compare results.

[kinghajj]

I'm not OCing the card, which I just bought a couple weeks ago just for mining, so cooling shouldn't be a problem. My temps usually range from 70-75 C. And fixing the problem is as simple as killing python2 and starting it again. I was thinking it could be a pool problem of some sort; I'm using slush's now, but don't remember ever having this problem when running in deepbit. I guess I could try the regular poclbm on slush's, and if the problem happens again, I'll try poclbm/-mod on deepbit and see what happens there.

[[Tycho]]

I'm not OCing the card, which I just bought a couple weeks ago just for mining, so cooling shouldn't be a problem. My temps usually range from 70-75 C. And fixing the problem is as simple as killing python2 and starting it again. I was thinking it could be a pool problem of some sort; I'm using slush's now, but don't remember ever having this problem when running in deepbit. I guess I could try the regular poclbm on slush's, and if the problem happens again, I'll try poclbm/-mod on deepbit and see what happens there.

By the way, why do you want to use poclbm-mod instead of poclbm ?

[Reto]

poclbm-mod optimizes efficiency, meaning less asks you send to the server. However, since deepbit's server is strong, this is unnecessary.

My theory is that on larger pools poclbm-mod actually hinders you, because poclbm-mod checks the entire getwork for answers. This makes it so you lag behind other users, causing you to submit stale shares.

I switched back to "vanilla" poclbm, and my stale share problem is virtually nonexistent now.

[geebus]

poclbm-mod optimizes efficiency, meaning less asks you send to the server. However, since deepbit's server is strong, this is unnecessary.

My theory is that on larger pools poclbm-mod actually hinders you, because poclbm-mod checks the entire getwork for answers. This makes it so you lag behind other users, causing you to submit stale shares.

I switched back to "vanilla" poclbm, and my stale share problem is virtually nonexistent now.

It doesn't cause you to submit stale shares. It supports long polling just the same as core poclbm and will grab new work if and when the block changes.

If there are problems with poclbm-mod, I recommend you post on the forums at bitcoinpool.com and I will address the issues directly.

[Veldy]

I'm not OCing the card, which I just bought a couple weeks ago just for mining, so cooling shouldn't be a problem. My temps usually range from 70-75 C. And fixing the problem is as simple as killing python2 and starting it again. I was thinking it could be a pool problem of some sort; I'm using slush's now, but don't remember ever having this problem when running in deepbit. I guess I could try the regular poclbm on slush's, and if the problem happens again, I'll try poclbm/-mod on deepbit and see what happens there.

Slush is down.  "502 Bad Gateway".  Since they score based on uptime this is unacceptable (everybody should get scored equally for the round when it comes up which hurts the people who have been continuously mining and those who stopped, but they can take it out of their "profit" to payout to the miners who were no up full-time since it is a site error and clearly the server they are using is not robust (nginx/0.7.65 .... come on, 1.0 was released awhile ago!).

[LMGTFY]

(nginx/0.7.65 .... come on, 1.0 was released awhile ago!).

Well, five days ago. Seriously, though, many OSs don't run with the latest and greatest. Debian Lenny (old stable) is still using Nginx 0.6.32-3, and Debian Squeeze (current stable) is on 0.7.67.

Edit: This looks to be just a problem with the website: I've just started mining against Slush's pool and all seems fine.

[rezin777]

Slush is down.  "502 Bad Gateway".  Since they score based on uptime this is unacceptable (everybody should get scored equally for the round when it comes up which hurts the people who have been continuously mining and those who stopped, but they can take it out of their "profit" to payout to the miners who were no up full-time since it is a site error and clearly the server they are using is not robust (nginx/0.7.65 .... come on, 1.0 was released awhile ago!).

I would guess this is the same thing that happened before. Not really a big deal.

http://bitcointalk.org/index.php?topic=1976.msg77673#msg77673

[Veldy]

Slush is down.  "502 Bad Gateway".  Since they score based on uptime this is unacceptable (everybody should get scored equally for the round when it comes up which hurts the people who have been continuously mining and those who stopped, but they can take it out of their "profit" to payout to the miners who were no up full-time since it is a site error and clearly the server they are using is not robust (nginx/0.7.65 .... come on, 1.0 was released awhile ago!).

I would guess this is the same thing that happened before. Not really a big deal.

http://bitcointalk.org/index.php?topic=1976.msg77673#msg77673

In hind sight, it was just the web server and not the pool.  There should be monitoring software in place to watch bitcoind among other services and daemons; it is open source and not expensive to implement.  Also, the site should give a message other than 502 if something recurring like this goes down; simple matter.  Without access to the website, I cannot tell if the miner is processing good shares or stale shares [no idea if the pool is working correctly].  These are simple matters for hosting [no, I don't host a website, but I am very familiar with administering Linux/FreeBSD even though I am a software engineer (Microsoft products primarily) by trade].

If you look at http://bitcoinwatch.com/ right now, you will note that slush has almost zero contribution at the moment.  That suggests to me that something bigger is going on [are all shares processed today stale at this point?].