archive snowblossom slack dev
prev 2018-06-23 next
2018-06-23 00:42:18
Rotonen
i have accumulated over the years a set of pretty extensive infra of my own in the scope of how to servers and services, i got this
2018-06-23 06:13:48
Rotonen
@Clueless make the user you create in those a service user
2018-06-23 06:15:00
Rotonen
@Clueless or i suppose for most `--system` is enough, oh well
2018-06-23 06:15:04
Clueless
`id -u snowblossom &>/dev/null || useradd --home-dir "$snowblossom_home" --create-home --system snowblossom`
2018-06-23 06:15:12
Clueless
yup, --system is a system user. :)
2018-06-23 06:15:46
Rotonen
what i mean is put selinux and polkit rules in place, but those are off per default in ubuntu?
2018-06-23 06:16:15
Clueless
I'm unfamiliar with selinux, but it may be worth something exploring for me.
2018-06-23 06:16:36
Clueless
I just compartmentalize my applications and servers enough that selinux wouldn't really help me anywhere.
2018-06-23 06:18:46
Rotonen
first steps would be to whitelist syscalls, resources, set up proper ACL, but i suppose enterprisey requirements are not for most
2018-06-23 06:22:51
Clueless
So I had installation scripts up pretty quickly, but for linux, they did things differently. So I was screwing with tuning them to work well with the deployment jars, and then again to do the build from source thing.
2018-06-23 06:23:35
Clueless
seems kinda stupid, but I made it pretty *global* $snowblossom_home/{node.sh, node.bat} is how you run ALL commands now.
2018-06-23 06:26:43
Fireduck
@Clueless you might consider restarting your pool with the changes I made toda
2018-06-23 06:26:45
Fireduck
today
2018-06-23 06:26:53
Fireduck
it should make the miner reported rates a bit more stable
2018-06-23 06:26:58
Fireduck
but not a huge deal
2018-06-23 06:27:09
Clueless
Sweet. I'm about to test floating ip switching on the pool.
2018-06-23 06:27:31
Fireduck
So I wasn't at a keyboard so couldn't type before
2018-06-23 06:27:36
Clueless
I've two vms, snowday1, snowday2, and I can throw up more.
2018-06-23 06:27:51
Fireduck
I think just a single vm is fine
2018-06-23 06:28:00
Fireduck
you are going to get a lot of uptime with just that
2018-06-23 06:28:10
Fireduck
and actually risk more downtime by making things complicated
2018-06-23 06:28:50
Fireduck
if you needed more than 2.5 or 3 9s, sure, but you don't need that
2018-06-23 06:29:11
Fireduck
but is fun to screw with things, so have fun
2018-06-23 06:29:24
Clueless
yeah, that's basically it
2018-06-23 06:29:43
Fireduck
fair enough
2018-06-23 06:29:45
Fireduck
i need to get to bed
2018-06-23 06:39:34
Rotonen
is the pool miner client actually going to survive the pool just silently changing?
2018-06-23 06:54:35
Clueless
@Rotonen Well, I (maybe erroneously) assume that switching a floating ip to another machine would close all the TCP connections and cause all the miners to reset their current work So no, there's no smooth transition (yet)
2018-06-23 07:59:07
Rotonen
dunno, can be either way, try first
2018-06-23 07:59:16
Rotonen
and try with something trivial you wrote and netcat
2018-06-23 12:49:45
Rotonen
@Fireduck this would look rather promising for distributing signed deployables: https://cachix.org/ disclaimer, i know domen personally (and thus trust him on that one as the service is fresh)
2018-06-23 13:59:07
GitHub
[snowblossomcoin/snowblossom] #86 Fastfail
2018-06-23 14:46:43
Fireduck
@Tyler Boone please update wiki mining tuning page.
2018-06-23 14:47:05
Fireduck
Fast fail in master, and check my notes for accuracy
2018-06-23 14:48:27
Clueless
I updated release notes and added to wiki sidebar for better visibility
2018-06-23 14:48:48
Clueless
I'll check them occasionally, especially with releases.
2018-06-23 15:34:14
Fireduck
n34t
2018-06-23 15:38:57
Fireduck
@Clueless so grpc will reconnect on tcp break
2018-06-23 15:39:19
Fireduck
however, for a while the miner will assume its work unit is still valid and submit shares based on it
2018-06-23 15:39:23
Fireduck
which will be rejected
2018-06-23 15:39:34
Fireduck
after a while, that work unit gets stale, the miner will stop using it
2018-06-23 15:39:53
Fireduck
and then the miner will detect that as a stall and reconnect/resubscribe
2018-06-23 15:40:09
Fireduck
which will get it a new work unit that actually works with the pool
2018-06-23 15:40:19
Fireduck
so it is a few minutes of downtime for the miner
2018-06-23 15:40:40
Fireduck
it would be better to do the failover code properly in the miner and give multiple pool hosts
2018-06-23 16:00:57
Protovist
The miner could also try the resubscribe when the share is rejected instead of waiting for stale work.
2018-06-23 16:01:30
Fireduck
That would be smart
2018-06-23 16:01:42
Fireduck
Could also add a heart beat stream
2018-06-23 16:02:12
Protovist
I was noticing that after restarting the pool last night.