Same here.  Twice today, my farm has happily continued submitting shares to the renter's pool 2 hours after the rental should have ended, with the status listed as "rented - available in a few moments."  This is a deal-breaker until it gets fixed.

Feel free to post here if you've done business with me.

I'm seeing  a lot of this in cgminer for the clevermining pool: "Rejected untracked stratum share from Pool 1"

Have you given any thought to having multi-algorithm rigs?  I'd have to check if cgminer and vertminer will coexist on the same box if one of them is idle at any given time, but if so, it'd be handy to point cgminer at betarigs port X and vertminer at port Y, and then have the same rig be rentable for both.

If someone rents it for Vertcoin, it becomes unavailable for scrypt, port Y starts serving up shares, and vertminer goes to town.  And likewise for scrypt.

I don't think so.  We're talking about a whole bunch of GPUs that sit at 0 hardware errors all week and then BAM, 4000 in a matter of minutes on every single GPU.  Then it stops and no more errors.  It's happened on multiple rentals. I'm sure it's related to the renter's pool, but I have no idea what they could be doing to cause it.

I don't plan to do anything about it, as it doesn't actually hurt anything, it's just odd.  I thought maybe you'd seen it before.

Thanks for the quick answer.  One thing to consider about the folks who operate big farms... I have no idea how professional (in terms of IT skill) they are in general, but I have a bunch of monitoring set up, keep a couple rigs out of my rental farm (and on one of the profit switching pools) as backup capacity, and my advertised hash rate is a little below actual anyway. 

So although I get your concerns about reliability on a larger scale, it's actually the large scale that allows me to offer reliability because someone could steal two rigs from my house and I'd still meet my rental obligation.

Just food for thought. :-)

Anyone notice wafflepool's hash rate?  6GH and rapidly rising.  Curiouser and curiouser...

No, I have a couple that throw errors at stock clocks and low intensity, whereas 90% of my GPUs stick at 0 with the same config.  Imperfect silicon, I guess.  But mathematically, 10 errors a day have no tangible effect on performance, so it's not worth worrying about until the errors accelerate.  WU is the same for the problem cards as the rest, in other words.

I haven't read the whole thread, so maybe this has been covered, but I figured I'd mention it as another data point.

This service is simple in that it doesn't require a proxy pool between the rigs and the renters' pools, which is good, but it's apparently not suited to offering one big 10+ megahash farm because you'd need network access to every individual instance of cgminer.

If this problem were to be solved, I'd give it a go, but managing 10 individual rentals is a much bigger pain than one large one.

Fantastic service and well implemented, thanks for that!

The only weirdness I'm seeing, and I only see it with Betarigs (as opposed to middlecoin/wafflepool/etc) is that sometimes, every GPU in the farm rings up several thousand hardware errors all at once, then it stops and it's back to normal.

Any idea what causes this, and if I need to worry about it?

My watchdog script originally would reboot a rig if a GPU hit 100 hardware errors (I have a couple that throw 10 errors a day or so), but I had to disable that feature with betarigs. 

Depends how long ago you set it up and what your unexchanged balances were at the time of the switch.  You probably have to let them reach equilibruim before starting to compare.

I haven't done the math it'd take to really say "I made $x more or less load-balancing."  Just eyeballing it, they're all in the same ballpark, all have fairly frequent technical problems, are all unpredictable over short periods of time.  And, oddly, do NOT necessarily move in lockstep... presumably because of the highly variable lag between receiving a cgminer share and converting it to confirmed Bitcoins.

I just got tired of trying to figure out who was best and figured "Together they'll all average X BTC/Mh/day, keep all my miners busy all the time, and besides I have other shit to do."

As the owner of a brick-and-mortar company that does work for businesses, as opposed to individuals, and mostly decent-sized corporations, there is something to this big miner vs small miner thing.

If I were to start a pool, honestly, I'd probably design it to attract larger miners (10MH+) and discourage smaller ones.  Providing support to a small number of large customers is SO much easier and more lucrative than a ton of small ones.

Or both, for failover and redundancy.  I have my cgminer.conf set to send 30% each to Middlecoin East and West, 20% to Wafflepool, and 20% to Clevermining.

Which should ensure that no miner ever sits around waiting for work, two pools can simultaneously drop off the net without affecting more than a day's profits, etc.

Hardware stable and auto-healing: check.  Redundant pools: check.  Ongoing work: verifying deposits every couple days.  *feet up* *beer in hand*.

Just FYI, given the outage yesterday, y'all can guard against having your rigs completely offline while still using the specific servers you want when they're up.  

I'm in the US, so the below cgminer.conf snippet splits 80% of my power between US East and West, 20% to wafflepool, and falls back to non-US Middlecoin servers if the first three are all down.  You can play with the quota numbers to create whatever behavior you want (making Wafflepool a last resort, load-balancing evenly among everything, etc).  It pretty much makes it set-and-forget.

"pools" : [
       {
               "quota" : "40;stratum+tcp://useast.middlecoin.com:3333",
               "user" : "...",
               "pass" : "x"
       },
       {
               "quota" : "40;stratum+tcp://uswest.middlecoin.com:3333",
               "user" : "...",
               "pass" : "x"
       },
       {
               "quota" : "20;stratum+tcp://wafflepool.com:3333",
               "user" : "...",
               "pass" : "x"
       },
       {
               "quota" : "0;stratum+tcp://eu.middlecoin.com:3333",
               "user" : "...",
               "pass" : "x"
       },
       {
               "quota" : "0;stratum+tcp://amsterdam.middlecoin.com:3333",
               "user" : "...",
               "pass" : "x"
       },
       {
               "quota" : "0;stratum+tcp://asia.middlecoin.com:3333",
               "user" : "...",
               "pass" : "x",
       }
],
"load-balance" : true,
"failover-only" : true,

Oops, I misspoke.  xintensity and thread-concurrency aren't (I think) interrelated.  But no, I didn't specify either thread-concurrency or worksize and cgminer appears to have figured them to be 8192 for 280Xs and 8960 for 7950s, and 256 worksize for both.

The Powercolor 280X has a stock core clock of 1030Mhz.  Factory-overclocked, I guess.  Which is why I could only overclock them from 715Kh/s to 740.

Yeah, I only meant stock clocks on the 280Xs.  And the point of xintensity is that you don't use thread concurrency.  It's used as a multiple of shaders.

The 280x's (all Powercolor) settled out to 715 for GPU 0 (no idea why it's so finicky) and 740 for the rest, and that's undervolted to 1.1V.  Unspectacular, but respectable.

Weirdly, the same core clock that gives 740kh on GPUs 1 to 4 drops GPU 0 down into the 500's. It insists on the stock 1030Mhz *shrug*

Using Kalroth's cgminer, I'm getting 620 to 650 on both XFX Double D 7950's and Sapphire dual-x 7950's, depending on how finicky a specific card is.  630 seems to be average with stock voltage, core clock 1055, and memory clock at stock (sapphire) or 900Mhz (XFX).

"xintensity" : "4",
"gpu-powertune" : "20",
"gpu-threads" : "2",
"shaders" : "1792",
"gpu-fan" : "0-100",
"auto-fan" : true,
"temp-target" : "70",

Those last 3 lines are important, as the sapphire cards love to auto-throttle if they get too warm.  And my 280x's are even worse.