V0.12.0.x msghand

[UdjinM6]

Revert some changes 40149c4#diff-9a82240fe7dfe86564178691cc57f2f1R1485 and cherry-pick 351593b
Should help with #535

[UdjinM6]

before

% time     seconds  usecs/call     calls    errors syscall
------ ----------- ----------- --------- --------- ----------------
 89.04    0.003298           1      2476      1235 futex
  9.64    0.000357           0     45249           gettimeofday
  1.16    0.000043           0       480           sendto
  0.16    0.000006           0      1235           clock_gettime
  0.00    0.000000           0        32           write
------ ----------- ----------- --------- --------- ----------------
100.00    0.003704                 49472      1235 total

after

% time     seconds  usecs/call     calls    errors syscall
------ ----------- ----------- --------- --------- ----------------
  0.00    0.000000           0        36           write
  0.00    0.000000           0       200           sendto
  0.00    0.000000           0      2451           gettimeofday
  0.00    0.000000           0       243         9 futex
  0.00    0.000000           0         1           restart_syscall
------ ----------- ----------- --------- --------- ----------------
100.00    0.000000                  2931         9 total

still smth is broken somewhere but looks much better now imo

EDIT: though it might be just because of the node restart... monitoring it now

[minersday]

Compiling and testing on one of my live masternodes..

[minersday]

OK, now it's running live,
changed the clocksource back to tsc = default.
Everything seems to be fine currently, there are no
futex: (Connection timed out), messages in strace as of now, there are some futex resource unavailable from time to time, and futex ..... = 0. but we will see.

looking realtime graph and
CPU usage went from ~20% to ~5%.
Still monitoring memory if it's starts to do the chainsaw or is it stable for now...
Network packets also went down from 3500 to around 2000.

[minersday]

Current graphs, I'll keep monitoring the situation...

[minersday]

Current graphs, we will see in day or two what happens..

[minersday]

Here's a week graph, you can see the CPU usage is good, even it's sawing a little, memory is all over the place, same as disk usage is increasing and is sawing These has to do something with the masternode count?
going to compile the latest release and try that.

[minersday]

Newest 0.46 release, looking very good. Memory usage back down to normal no sawing and cpu-usage too (at least not yet), going to update the rest of the masternodes. Updated around 7:10


edit: and disk usage is back down to normal also, from daily graph

[UdjinM6]

@minersday thank you for extended testing and report! 👍

[schinzelh]

@minersday many thanks for your testing from me, too!

[minersday]

NP =)

[minersday]

Here's latest graphs for you please (daily range), you can clearly see at the end when updated to .46.

[schinzelh]

I can confirm this, i have updated some nodes and the load reduces significantly (at least after startup)