[2.2.0] RandomX algo sometimes doesn't init

[electroape]

Not sure if it's XMRigCC or upstream but sometimes when switching from other algo to RandomX (ARQ, not sure about Loki), it doesn't init how it's supposed to and prints this : [2019-11-10 00:46:00.613] cpu use profile rx/arq (4 threads) scratchpad 256 KB [2019-11-10 00:46:00.613] cpu READY threads 4(4) huge pages 4/4 100% memory 1024 KB (1 ms) instead of this : [2019-11-10 01:13:05.321] rx init dataset algo rx/arq (4 threads) seed 19a4fa4705a9788a... [2019-11-10 01:13:05.321] cpu use profile rx/arq (4 threads) scratchpad 256 KB [2019-11-10 01:13:05.321] rx #0 allocate 2336 MB (2080+256) for RandomX dataset & cache [2019-11-10 01:13:05.321] cpu READY threads 4(4) huge pages 2/4 50% memory 1024 KB (3 ms) [2019-11-10 01:13:05.336] rx #0 allocate done huge pages 0/1168 0% +JIT (11 ms) [2019-11-10 01:13:13.056] rx #0 init done (7743 ms) thus not mining. Restarting the miner helps.

[electroape]

Also, since RandomX allocates huge amounts of RAM it would be useful to see client's RAM amount in the dashboard.

[electroape]

Also, in some cases miner doesn't continue mining after a loss of connection.

[2019-11-10 12:06:00.941] accepted (41/0) diff 97158 (228 ms) [2019-11-10 12:17:40.449] [CC-Client] error: unable to performRequest POST -> http://x/client/setClientStatus?clientId=x [2019-11-10 12:17:43.513] speed 10s/60s/15m n/a n/a 3358.7 H/s max 3675.6 H/s [2019-11-10 12:17:47.185] [arqma.herominers.com:10640] connect error: "invalid state: 5" [2019-11-10 12:17:50.449] [CC-Client] error: unable to performRequest POST -> http://x/client/setClientStatus?clientId=x [2019-11-10 12:17:51.357] use pool arqma.herominers.com:10640 195.201.104.207 [2019-11-10 12:17:51.357] new job from arqma.herominers.com:10640 diff 15000 algo rx/arq height 304386 [2019-11-10 12:18:43.790] speed 10s/60s/15m n/a n/a 2728.1 H/s max 3675.6 H/s [2019-11-10 12:19:44.438] speed 10s/60s/15m n/a n/a 2679.5 H/s max 3675.6 H/s [2019-11-10 12:20:45.094] speed 10s/60s/15m n/a n/a 1130.5 H/s max 3675.6 H/s [2019-11-10 12:21:45.293] speed 10s/60s/15m n/a n/a 125.8 H/s max 3675.6 H/s [2019-11-10 12:22:45.333] speed 10s/60s/15m n/a n/a n/a H/s max 3675.6 H/s [2019-11-10 12:23:45.380] speed 10s/60s/15m n/a n/a n/a H/s max 3675.6 H/s [2019-11-10 12:24:45.418] speed 10s/60s/15m n/a n/a n/a H/s max 3675.6 H/s [2019-11-10 12:25:45.456] speed 10s/60s/15m n/a n/a n/a H/s max 3675.6 H/s [2019-11-10 12:26:45.494] speed 10s/60s/15m n/a n/a n/a H/s max 3675.6 H/s [2019-11-10 12:27:45.532] speed 10s/60s/15m n/a n/a 0.0 H/s max 3675.6 H/s [2019-11-10 12:28:45.573] speed 10s/60s/15m n/a n/a 0.0 H/s max 3675.6 H/s [2019-11-10 12:29:45.612] speed 10s/60s/15m n/a n/a 0.0 H/s max 3675.6 H/s [2019-11-10 12:30:45.665] speed 10s/60s/15m n/a n/a 0.1 H/s max 3675.6 H/s [2019-11-10 12:31:45.702] speed 10s/60s/15m n/a n/a 0.1 H/s max 3675.6 H/s [2019-11-10 12:32:45.748] speed 10s/60s/15m n/a n/a 0.1 H/s max 3675.6 H/s [2019-11-10 12:33:45.781] speed 10s/60s/15m n/a n/a 0.1 H/s max 3675.6 H/s

[electroape]

In some as in about 50% of my ARQ miners are doing nothing right now after temporary connection loss. I suspect it has smth to do with memory heavy algo itself, smth with memory being reclaimed already and it can't use it anymore i dunno. Maybe there should be some re-init in that case ?

[electroape]

Same behaviour after dev donate : [2019-11-10 15:50:37.120] accepted (167/0) diff 80992 (115 ms) [2019-11-10 15:50:40.998] accepted (168/0) diff 80992 (111 ms) [2019-11-10 15:51:09.029] speed 10s/60s/15m 4520.2 4592.9 4506.6 H/s max 5049.8 H/s [2019-11-10 15:51:11.228] new job from arqma.herominers.com:10640 diff 161988 algo rx/arq height 304493 [2019-11-10 15:51:13.715] dev donate started [2019-11-10 15:51:13.715] new job from donate.graef.in:443 diff 2048148 algo cn-pico [2019-11-10 15:51:13.730] cpu use profile cn-pico (4 threads) scratchpad 256 KB [2019-11-10 15:51:13.777] cpu READY threads 4(8) huge pages 3/4 75% memory 2048 KB (18 ms) [2019-11-10 15:51:22.106] new job from donate.graef.in:443 diff 1024074 algo cn-pico [2019-11-10 15:52:09.087] speed 10s/60s/15m 2905.1 n/a n/a H/s max 2984.3 H/s [2019-11-10 15:52:13.705] dev donate finished [2019-11-10 15:52:13.705] new job from arqma.herominers.com:10640 diff 53995 algo rx/arq height 304493 [2019-11-10 15:52:13.705] cpu use profile rx/arq (4 threads) scratchpad 256 KB [2019-11-10 15:52:13.752] cpu READY threads 4(4) huge pages 3/4 75% memory 1024 KB (3 ms) [2019-11-10 15:52:57.759] new job from arqma.herominers.com:10640 diff 53995 algo rx/arq height 304494 [2019-11-10 15:53:11.228] new job from arqma.herominers.com:10640 diff 26998 algo rx/arq height 304494 [2019-11-10 15:53:18.911] speed 10s/60s/15m n/a n/a n/a H/s max 5049.8 H/s [2019-11-10 15:54:11.229] new job from arqma.herominers.com:10640 diff 15000 algo rx/arq height 304494 [2019-11-10 15:54:18.949] speed 10s/60s/15m n/a n/a n/a H/s max 5049.8 H/s [2019-11-10 15:55:18.978] speed 10s/60s/15m n/a n/a n/a H/s max 5049.8 H/s [2019-11-10 15:55:23.930] new job from arqma.herominers.com:10640 diff 15000 algo rx/arq height 304495 [2019-11-10 15:56:22.467] speed 10s/60s/15m n/a n/a n/a H/s max 5049.8 H/s [2019-11-10 15:57:22.499] speed 10s/60s/15m n/a n/a n/a H/s max 5049.8 H/s [2019-11-10 15:58:22.535] speed 10s/60s/15m n/a n/a n/a H/s max 5049.8 H/s [2019-11-10 15:59:22.772] speed 10s/60s/15m n/a 0.6 n/a H/s max 5049.8 H/s [2019-11-10 15:59:36.889] new job from arqma.herominers.com:10640 diff 15000 algo rx/arq height 304496

[Bendr0id]

This sounds like an upstream problem to me. I tried to reproduce it but i wasn't able to.

It looks like the miner is still receiving new jobs, but for whatever reason its not able to start a worker on it..

[Bendr0id]

Is this still happening?

[electroape]

Yes, it's just that i've moved all low-RAM miners to Extremelite instead. The issue in first post seems different tho (different log output like not for Rx init). I'll try to reproduct it with 2.2.1 now.

[electroape]

I didn't see that (improper init) happening on miners that mine Rx without problems (those with 8GB RAM or more) so i guess it might be related too.

[electroape]

Of, behaviour described in the first post is just my misunderstanding, it doesn't print out 'rx init' etc entries if it did init it already and miner didn't restart. But low-RAM miners still don't mine Rx even though there should be supposedly low-RAM slow mode. Sometimes they do mine after restart but if they temporarily lose connection or after dev donate with different algo, they might freeze anyway.

[Bendr0id]

Is this issue still present?