From: Agata Kruszona-Z. <ch...@mo...> - 2020-01-28 12:45:26
|
> From: *Jay Livens* <jl...@sl... <mailto:jl...@sl...>> > Date: Tue, Jan 21, 2020 at 2:12 AM > > > Hi, > > Every now and then, I need to restart the mfsmaster service for reasons > such as upgrading MooseFS or perhaps a reboot on the MFS node stemming > from a kernel upgrade. My system only has one mfsmaster instance and so > failing over is not an option. This is in a home environment and so the > traffic on my MooseFS instance is low; however, I want to be sure that I > restart MooseFS in the safest way. I have tried the following: > > 1. Stop all chunskservers (cgiservers and metaloggers) and then the > master. Start the master service and then start all the slaves. (I > have automated this with Ansible, but it is still a bit bothersome.) > > 2. Just bounce the mfsmaster processes (mfsmaster, mfscgi > and mfschunkerser) and leaving all chunkservers up. This worked, but I > worry that doing so could cause corruption issues. (It did not seem to > when I tried it.) > > Anyway, I am trying to get the perspectives of the experts here and > understand how you folks handle this. > > TIA Hi, If you only want to restart your master, you don't need to stop everything. To be on the safe side, you should stop any client processes that write data to your MooseFS instance (via mfsmount) before you restart your master, but it's not necessary to stop chunkservers. -- Agata Kruszona-Zawadzka MooseFS Team |