Watchdog causes repeated miner restarts

[use-dashes-instead]

Windows compile of mm 1.5. Metaminer is getting into a loop restarting the miner when no results are being returned.

1. The watchdog is set at 600s (ten minutes).
2. If there are no results accepted by the pool in ten minutes, mm restarts the miner.
3. After the initial restart, mm lets the miner run for about a minute and, then, restarts it again due to no results for the last 600 seconds. After another minute, it starts again.
4. The miner restarts repeat until the miner manages to return a result during its short run time, or mm is manually restarted.

[use-dashes-instead]

Possible fix: Watchdog timer resets when mm restarts the miner, irrelevant of the reason for the restart.

Part of the issue seems to be that mm is counting the time before a miner restart, whether for change of algorithm or no results during set period. This would allow the miner the full time to find a result, rather than constantly restarting the miner after a short period.

[Spudz76]

five minutes between results? noooooooooooo

I disagree that 600s is short, it's about 8x farther out than the real limit, to catch runwaway zombies, you should target a block time of around 90s maximum.

600s is the default "hang up this nerd fell alseep" timeout on TCP also so all kind of network devices between you and the pool got tired of waiting for your slow rig. :)

You need to run against a port with a difficulty of around your_hashrate * 90
Examples: some 25000 diff port needs 25000/90= 277.7H/s in order to deliver 25000 results within 90s

If your hashrate is like 30 and you're on a 25000 port it will take EIGHT-HUNDRED THIRTY-THREE SECONDS which is like nobody is even going to know about that job anymore and can't pay you, if you were getting any results at all they were literally lucky (early).