OFFICIAL CGMINER mining software thread for linux/win/osx/mips/arm/r-pi 4.11.0 - page 435.

The00Dustin

hero member

Activity: 807

Merit: 500

Quote from: -ck on November 26, 2012, 06:40:41 PM

Debug build would be helpful. If you're running fedora that suggests you're building it yourself. If that's the case, add "-g" to your CFLAGS, rebuild your cgminer 2.9.5 (without stripping the file), enable coredumps with "ulimit -c unlimited" and then run whatever it is that's crashing. Once you get a "core dumped" message at the end of running it, then run:
gdb cgminer core
bt full

And post the information from that please.

I'm getting "No stack." I assume that makes it obvious I've never done this before...

ETA:

Code:

cat /usr/src/cgminer-2.9.5/Makefile | grep CFLAGS\ =
CFLAGS = -I/usr/src/ati-stream-sdk-v2.1-lnx64/include -g

Luke-Jr

legendary

Activity: 2576

Merit: 1186

Quote from: -ck on November 26, 2012, 06:25:28 PM

Throwing out work with every notify message on stratum, which comes on average every 30 seconds, would have the same effect as a longpoll every 30 seconds. That's a lot of work to discard midstream. Why on earth would there even be a clean_work method with stratum if every notify implied we throw out the work?

You're not supposed to throw out the shares on LPs. Restarting work has no cost except on BFLs (where you have a bug that loses valid shares).

Quote from: Askit2 on November 26, 2012, 06:30:33 PM

So Junior is saying my reject ~9minutes into the block is because CGMiner didn't change to new work fast enough?

That sounds highly unlikely, and unrelated. What kind of a reject?

TheHarbinger

sr. member

Activity: 378

Merit: 250

Why is it so damn hot in here?

I have learned quite a bit since starting with BitCoin, I will try to sum it all up for anyone who doesn't want to read every post in these forums.

Luke Jr. is never wrong.
Inaba is never wrong.
RealSolid is never wrong.

I can only wait in eager anticipation for the day when EMC adds SC mining, with SC mining added to BFG miner. Then we will truly be living in a utopia!

Askit2

hero member

Activity: 981

Merit: 500

DIV - Your "Virtual Life" Secured and Decentralize

Quote from: Inaba on November 26, 2012, 06:32:50 PM

Are you sure it's not an issue of share leakage?

It's a tad hard to discern. It could be work that got misrouted to us1 pool.... Possibly there is an easy way to tell I haven't noticed in the past?
It happens from time to time. I do agree the other issue with stratum (disconncts lose shares) is a larger loss for me.

-ck

legendary

Activity: 4088

Merit: 1631

Ruu \o/

Quote from: The00Dustin on November 26, 2012, 06:36:42 PM

I hate to interrupt all of this stratum discussion, but I am having crashes with GBT. Running Fedora 15 with SDK 2.1 I think. 2.9.4 would exit with something like error code 11 (or 15 maybe?), and 2.9.5 segfaults. I don't need GBT, so I should probably try to remember to try --fix-protocol instead of switching back to 2.8.7, but for now, I'm on 2.8.7 again. I didn't report the issue with 2.9.4, but I suppose I should've.
ETA: con, note that I wouldn't be against GBT being scrapped/removed from cgminer based on anything I've seen, but from a userbase standpoint, that might not be the option. I'm only providing this most basic info in case it is helpful in some way.

Debug build would be helpful. If you're running fedora that suggests you're building it yourself. If that's the case, add "-g" to your CFLAGS, rebuild your cgminer 2.9.5 (without stripping the file), enable coredumps with "ulimit -c unlimited" and then run whatever it is that's crashing. Once you get a "core dumped" message at the end of running it, then run:
gdb cgminer core
bt full

And post the information from that please.

The00Dustin

hero member

Activity: 807

Merit: 500

I hate to interrupt all of this stratum discussion, but I am having crashes with GBT. Running Fedora 15 with SDK 2.1 I think. 2.9.4 would exit with something like error code 11 (or 15 maybe?), and 2.9.5 segfaults. I don't need GBT, so I should probably try to remember to try --fix-protocol instead of switching back to 2.8.7, but for now, I'm on 2.8.7 again. I didn't report the issue with 2.9.4, but I suppose I should've.
ETA: con, note that I wouldn't be against GBT being scrapped/removed from cgminer based on anything I've seen, but from a userbase standpoint, that might not be the option. I'm only providing this most basic info in case it is helpful in some way.

Inaba

legendary

Activity: 1260

Merit: 1000

Are you sure it's not an issue of share leakage?

Askit2

hero member

Activity: 981

Merit: 500

DIV - Your "Virtual Life" Secured and Decentralize

Quote from: Luke-Jr on November 26, 2012, 06:11:36 PM

Quote from: -ck on November 26, 2012, 05:58:15 PM

Okay I see the problem here on EMC stratum.

Based on a commit luke-jr did to bfgminer, it seems he thinks that new work notification by stratum mandates that all work be thrown out in favour of the new work, because he FORCES the clean flag. However no other pool actually expects this, only forcing a clean when they actually send the work clean message. So I'm going to go out on a limb here and say the problem lies with the implementation of stratum on EMC, as coded up by luke-jr.

No, I think new work notification should be used as soon as possible without throwing out the previous jobs, just like it is for getwork and GBT (which have an equivalent submitold indicator). This is implied by the current Stratum documentation:

Quote

clean_jobs - When true, server indicates that submitting shares from previous jobs don't have a sense and such shares will be rejected. When this flag is set, miner should also drop all previous jobs, so job_ids can be eventually rotated.

Admittedly, there is room for improvement in stratum here, since it could support 3 states:

Previous jobs are invalid, don't send shares (getwork/GBT submitold=false; impossible with cgminer now)
Previous jobs are valid, but please start using this immediately (getwork/GBT submitold=true; how cgminer interprets clean_jobs=false now)
Previous jobs are valid; use this for new work at your convenience (how cgminer is interpreting clean_jobs=true now; no getwork/GBT equivalent)

Perhaps slush could clarify the current meaning, but it would be disappointing to learn stratum discards another fix for a problem we had already solved with getwork/GBT. Regardless of the current meaning, I intend to suggest the tristate when stratum's BIP process begins.

So Junior is saying my reject ~9minutes into the block is because CGMiner didn't change to new work fast enough?
Wow that seems unlikely. Apparenly EMC pays for invalids now for ~9 minutes. Or there is an issue more like Con said.
If I made a bet on this I would guess I had a legitimate share that will now not get paid. No disconect (unless it is hidden on the primary pool) (not debug, verbose or rpc debug). Not hold over stale (expiry isn't that high any more). It shouldn't be working on old work UNLESS I am paid virtually without punishment (Very doubtful).

-ck

legendary

Activity: 4088

Merit: 1631

Ruu \o/

Throwing out work with every notify message on stratum, which comes on average every 30 seconds, would have the same effect as a longpoll every 30 seconds. That's a lot of work to discard midstream. Why on earth would there even be a clean_work method with stratum if every notify implied we throw out the work?

kano

legendary

Activity: 4634

Merit: 1851

Linux since 1997 RedHat 4

Quote from: Luke-Jr on November 26, 2012, 06:11:36 PM

Quote from: -ck on November 26, 2012, 05:58:15 PM

Okay I see the problem here on EMC stratum.

Based on a commit luke-jr did to bfgminer, it seems he thinks that new work notification by stratum mandates that all work be thrown out in favour of the new work, because he FORCES the clean flag. However no other pool actually expects this, only forcing a clean when they actually send the work clean message. So I'm going to go out on a limb here and say the problem lies with the implementation of stratum on EMC, as coded up by luke-jr.

No, I think new work notification should be used as soon as possible without throwing out the previous jobs, just like it is for getwork and GBT (which have an equivalent submitold indicator). This is implied by the current Stratum documentation:

Quote

clean_jobs - When true, server indicates that submitting shares from previous jobs don't have a sense and such shares will be rejected. When this flag is set, miner should also drop all previous jobs, so job_ids can be eventually rotated.

Admittedly, there is room for improvement in stratum here, since it could support 3 states:

Previous jobs are invalid, don't send shares (getwork/GBT submitold=false; impossible with cgminer now)
Previous jobs are valid, but please start using this immediately (getwork/GBT submitold=true; how cgminer interprets clean_jobs=false now)
Previous jobs are valid; use this for new work at your convenience (how cgminer is interpreting clean_jobs=true now; no getwork/GBT equivalent)

Perhaps slush could clarify the current meaning, but it would be disappointing to learn stratum discards another fix for a problem we had already solved with getwork/GBT. Regardless of the current meaning, I intend to suggest the tristate when stratum's BIP process begins.

Ah yes lets just ignore the GBT spec while we're at it and solve that piece of crap and do it properly

You changed your code and ignored the spec ... good idea that one Tongue

Luke-Jr

legendary

Activity: 2576

Merit: 1186

Quote from: -ck on November 26, 2012, 05:58:15 PM

Okay I see the problem here on EMC stratum.

Based on a commit luke-jr did to bfgminer, it seems he thinks that new work notification by stratum mandates that all work be thrown out in favour of the new work, because he FORCES the clean flag. However no other pool actually expects this, only forcing a clean when they actually send the work clean message. So I'm going to go out on a limb here and say the problem lies with the implementation of stratum on EMC, as coded up by luke-jr.

No, I think new work notification should be used as soon as possible without throwing out the previous jobs, just like it is for getwork and GBT (which have an equivalent submitold indicator). This is implied by the current Stratum documentation:

Quote

clean_jobs - When true, server indicates that submitting shares from previous jobs don't have a sense and such shares will be rejected. When this flag is set, miner should also drop all previous jobs, so job_ids can be eventually rotated.

Admittedly, there is room for improvement in stratum here, since it could support 3 states:

Previous jobs are invalid, don't send shares (getwork/GBT submitold=false; impossible with cgminer now)
Previous jobs are valid, but please start using this immediately (getwork/GBT submitold=true; how cgminer interprets clean_jobs=false now)
Previous jobs are valid; use this for new work at your convenience (how cgminer is interpreting clean_jobs=true now; no getwork/GBT equivalent)

Perhaps slush could clarify the current meaning, but it would be disappointing to learn stratum discards another fix for a problem we had already solved with getwork/GBT. Regardless of the current meaning, I intend to suggest the tristate when stratum's BIP process begins.

-ck

legendary

Activity: 4088

Merit: 1631

Ruu \o/

Okay I see the problem here on EMC stratum.

Based on a commit luke-jr did to bfgminer, it seems he thinks that new work notification by stratum mandates that all work be thrown out in favour of the new work, because he FORCES the clean flag. However no other pool actually expects this, only forcing a clean when they actually send the work clean message. So I'm going to go out on a limb here and say the problem lies with the implementation of stratum on EMC, as coded up by luke-jr.

-ck

legendary

Activity: 4088

Merit: 1631

Ruu \o/

There's something else about stratum on EMC, and that appears to be lots of disconnects? That's another limitation of stratum, that once you disconnect the shares cannot be submitted on reconnect. Slush is going to have to add a reconnect option to it, or any disconnect of the socket will always be associated with shares lost if you're actively mining on it. Certainly it appears to be a combination of things and not just the frequent retargetting, although that is easy to see - the reject will say something like "share above target". "Unknown work" after longpoll is a true stale, work from the previous block. Not sure what's going on with GBT on EMC, but yes the rejects appear higher there too. For the time being you can still mine on getwork with --fix-protocol, but I'd rather see the pool issues sorted out.

Askit2

hero member

Activity: 981

Merit: 500

DIV - Your "Virtual Life" Secured and Decentralize

My EMC rejects are 1.2% with stratum. I understand the post LP dump and I appreciate the chance at a paid share. The intermediate Rejected [Blah blah blah] (Unknown-Work). Usually this is in the middle of 2 other results found in a work unit or at least the 2nd submission at a time. Sometimes its first, sometimes last I guess.

It doesn't seem like unknown work. Things before it get accepted, after it also accepted. If the work was actually unknown wouldn't I expect more then a single reject in a batch of submissions? I suppose I could be piling up a result every so often that isn't accepted and on resubmit it is rejected. Expiry is likely a little high at 120s but my lower results are running the same.

Setup 1 BFL Single and CGMiner 2.9.5.

Before junior pipes up about GBT. I ran .2% rejects on GW, .8% on GBT, 1.2% on stratum on eclipse. All results on eclipse. The reduction of network traffic is far nicer then the slight reduction in rate.

Also on Stratum I am seeing a 856.X overall hashrate, Last week I managed 852.X with GBT and more network load. This could be new difficulty related or luck as it has only been on for 26K shares.

kano

legendary

Activity: 4634

Merit: 1851

Linux since 1997 RedHat 4

Quote from: P_Shep on November 26, 2012, 04:11:01 PM

Quote from: Luke-Jr on November 26, 2012, 04:06:44 PM

Quote from: -ck on November 26, 2012, 03:47:16 PM

Quote from: P_Shep on November 26, 2012, 01:19:21 PM

So the issue with stales on stratum EMC...
Are they actually stale, or is it submitting shares which are below the expected target or something?

I'm getting 1% stale on EMC, and 0.1 on Ozcoin.

This is the issue I've been discussing at length on this thread with Inaba. Yes it's just submitting shares below target after diff rises.

That wouldn't make them stale... and EMC accepts shares at the lower target for work issued before it changed...

I'm miss-labelling here, sorry. I didn't mean stale, I meant rejected (for whatever reason).

Yes we all understood that ... except one 'person' Tongue

P_Shep

legendary

Activity: 1795

Merit: 1208

This is not OK.

Quote from: Luke-Jr on November 26, 2012, 04:06:44 PM

Quote from: -ck on November 26, 2012, 03:47:16 PM

Quote from: P_Shep on November 26, 2012, 01:19:21 PM

So the issue with stales on stratum EMC...
Are they actually stale, or is it submitting shares which are below the expected target or something?

I'm getting 1% stale on EMC, and 0.1 on Ozcoin.

This is the issue I've been discussing at length on this thread with Inaba. Yes it's just submitting shares below target after diff rises.

That wouldn't make them stale... and EMC accepts shares at the lower target for work issued before it changed...

I'm miss-labelling here, sorry. I didn't mean stale, I meant rejected (for whatever reason).

juhakall

sr. member

Activity: 658

Merit: 250

The share rejection problem I have on EMC (described a few posts ago) isn't connected to stratum. I'm using http://us2.eclipsemc.com:8337 as the URL and miner.php says it's not using stratum. I'm not sure if it's using GBT, though. How do I check that?

Luke-Jr

legendary

Activity: 2576

Merit: 1186

Quote from: -ck on November 26, 2012, 03:47:16 PM

Quote from: P_Shep on November 26, 2012, 01:19:21 PM

So the issue with stales on stratum EMC...
Are they actually stale, or is it submitting shares which are below the expected target or something?

I'm getting 1% stale on EMC, and 0.1 on Ozcoin.

This is the issue I've been discussing at length on this thread with Inaba. Yes it's just submitting shares below target after diff rises.

That wouldn't make them stale... and EMC accepts shares at the lower target for work issued before it changed...

-ck

legendary

Activity: 4088

Merit: 1631

Ruu \o/

Quote from: P_Shep on November 26, 2012, 01:19:21 PM

So the issue with stales on stratum EMC...
Are they actually stale, or is it submitting shares which are below the expected target or something?

I'm getting 1% stale on EMC, and 0.1 on Ozcoin.

This is the issue I've been discussing at length on this thread with Inaba. Yes it's just submitting shares below target after diff rises.

Try the EMC GBT server until it's resolved.

Lem

newbie

Activity: 78

Merit: 0

Quote from: juhakall on November 26, 2012, 02:26:09 PM

I removed failover-only from my config because share leakage is lower

I did the same. I can confirm that with 2.9.5 share leakage has actually gone back down to a negligible minimum.

Well done!

Topic: OFFICIAL CGMINER mining software thread for linux/win/osx/mips/arm/r-pi 4.11.0 - page 435. (Read 5805983 times)