Re: [9front] rimport error: tlsclient: auth_proxy: fauth_proxy start: too much activity

Re: [9front] rimport error: tlsclient: auth_proxy: fauth_proxy start: too much activity

[cinap_lenrek]

too many auth procs spawned in factotum. it bounds the number of
processes to 16 and refuses to start another one. see:

/sys/src/cd/auth/factotum/fs.c:493:

static void
fsrpcio(Req *r)
{
	Fsstate *fss;
	Srv *srv;

	fss = r->fid->aux;
	srv = r->srv;
	if(srv->sref.ref == 1 && srv->rref.ref > 16){
		respond(r, "too much activity");
		return;
	}
	srvrelease(srv);
	qlock(fss);
	if(r->ifcall.type == Tread)
		rpcread(r);
	else
		rpcwrite(r);
	qunlock(fss);
	srvacquire(srv);
}

my guess is that factotum acts as a client and the auth server is
slow to respond? or theres a network issue?

factotum should eventually timeout the authserver communication.

--
cinap

Re: [9front] rimport error: tlsclient: auth_proxy: fauth_proxy start: too much activity

[Stanley Lieber]

On Mar 30, 2018, at 11:15 PM, cinap_lenrek@felloff.net wrote:
> 
> my guess is that factotum acts as a client and the auth server is
> slow to respond? or theres a network issue?
> 
> factotum should eventually timeout the authserver communication.

The error shows up in my logging for my backup scripts, which runs a new rimport each time the script is run (several times per hour). It’s definitely plausible that there could be 16 or more firing off at more or less the same time, all backed by the same factotum acting as a client.

However, the backup script runs on the auth server, shared by all the machines involved.

sl

Re: [9front] rimport error: tlsclient: auth_proxy: fauth_proxy start: too much activity

[cinap_lenrek]

you can try increasing the limit. the reason it spawns processes is that when
one authserver is broken/dead, it wont hang the whole factotum. on the other
hand, we dont want it to forkbomb as well so it can recover once the authserver
is restored.

another approach is to revert to old behaviour once the limit is reached. so
it wont lock up unless you try hard. and maybe then you deserve it?

you could also try to avoid the issue by staggering your cronjobs, so they
wont all hammer in at the same time.

--
cinap