When using glusterfs in combination with NFS mounts gluster(fs)?d should start after rpcbind in order to correctly register with rpcbind. I debated whether this should be need but the reality is that most people probably will not be using glusterfs via nfs, and just adding rpcstat to the default runlevel should be fine. Reproducible: Always
Yeah, should be handled by users imho, thx mate
I am initializing sys-cluster/glusterfs-6.5 according to https://docs.gluster.org/en/latest/Quick-Start-Guide/Quickstart/ The command systemctl enable glusterd;systemctl start glusterd just printed me Failed to start glusterd.service: Unit rpcbind.service not found. So I assume it would be wise to add net-nds/rpcbind to RDEPEND. Please reopen this bug until the missing dependency is resolved. -- Portage 2.3.69 (python 3.6.9-final-0, default/linux/amd64/17.0/desktop, gcc-8.2.0, glibc-2.29-r2, 4.20.0-sabayon x86_64) ================================================================= System uname: Linux-4.20.0-sabayon-x86_64-Intel-R-_Core-TM-_i5-9500_CPU_@_3.00GHz-with-gentoo-2.6 KiB Mem: 16190320 total, 519808 free KiB Swap: 0 total, 0 free Timestamp of repository gentoo: Thu, 05 Dec 2019 00:45:01 +0000 Head commit of repository gentoo: dc9289a782f74c49bf8f248425e17cc2bc672d7a sh bash 4.4_p23-r1 ld GNU ld (Gentoo 2.30 p5) 2.30.0 app-shells/bash: 4.4_p23-r1::gentoo
rpcbind is only required if you're using NFS exports. Which I highly discourage.
I cannot even start glusterd even if I don't want to use NFS exports, systemctl won't let me start glusterd. Maybe the good solution would be to add a USE=-nfs to the package, and to add a dependency on nfs for rpcbind.
Also, it would be wise to set -ipv6 by default: see last post of https://forums.gentoo.org/viewtopic-t-1082454-start-0.html and I did reproduce the problem and this fix solved my problem.
(In reply to Daa Jaa from comment #5) > Also, it would be wise to set -ipv6 by default: see last post of > https://forums.gentoo.org/viewtopic-t-1082454-start-0.html > and I did reproduce the problem and this fix solved my problem. I concur. I've never had success with glusterfs with +ipv6. (In reply to Daa Jaa from comment #4) > I cannot even start glusterd even if I don't want to use NFS exports, > systemctl won't let me start glusterd. > > Maybe the good solution would be to add a USE=-nfs to the package, and to > add a dependency on nfs for rpcbind. This is more tricky that you'd imagine. The issue is that new volumes by default has NFS enabled. Which is why I requested for the "after rpcbind" in openrc init script. Thus I could manually add/remove rpcbind to the runlevels as required. I'm not sure how systemd deals with same. Don't particularly care either. Current status quo works for me.
