Forum/Site Reboot Policy /Advice

[aegersz]

hello smf gurus,

i have been rebooting my forum every week BUT should i bother or what ?

what do you say about system reboots (mine is CentOS 6.7), please ?

[Illori]

you mean server reboots? why do you feel like you need to reboot weekly?

[aegersz]

yes, server reboots.

as a mainframe guy, we always re-IPLed (reboot equivalent) daily, to refresh the system's control blocks.

i guess i am living in the past ... but i thought that it might release some memory and page space ...

i also reboot the server to make sure that it bounces back up ... after modifying IPTABLES or using CHKCONFIG etc.

but shall i bother ? i haven't rebooted for a week and wanted to let it run.

[aegersz]

i guess your answer is don't bother rebooting the server for the forum's sake.

ok, i will let it run

[青山 素子]

as a mainframe guy, we always re-IPLed (reboot equivalent) daily, to refresh the system's control blocks.

Really? Everything I've heard is that mainframes (like the RS6000 and similar) aren't usually restarted often because of concerns over downtime. There are plenty of stories over the uptime reliability of Tandem, VAX, and similar systems both minicomputer and mainframes.

i guess i am living in the past ... but i thought that it might release some memory and page space ...

It won't really. Modern systems are usually very good about memory usage and clearing things out. For Linux at least, swap usage will always show used once it has been marked "dirty", even if that space is marked for re-use. Some poorly-coded applications may have issues, but nearly all popular ones have had any issues found and fixed by now.

i also reboot the server to make sure that it bounces back up ... after modifying IPTABLES or using CHKCONFIG etc.

Generally a good idea, but you can group those checks. Also, you probably shouldn't have so much configuration churn on a production system.

but shall i bother ? i haven't rebooted for a week and wanted to let it run.

I generally only reboot during kernel or major system library updates. Sometimes it's 30 or more days, sometimes less.

[aegersz]

my mainframes were IBM 370/390/ ESA large systems ... 24 and 31-bit.

thanks for the thorough answers - this is my first personal production Linux server and still settling in as I tweak this and that.

it seems stable and very quick now.

ok, let's see if she can stay up for a month - that'd be cool.

pardon my 1960's - 2000's technology experience - now all but redundant now ... so I'm unemployed and serving a drug harm reduction and informational forum for kicks !

have got over 550 members, mostly in the last few months so the site has to rock.

[青山 素子]

My back-end monitoring and CI server and front-end load balancer at work are currently up 41 days. They'll be rebooted soon as there's a major system update pending a maintenance period.

I've had servers up much longer due to client requirements. One of the longest ones I've been responsible for was over 3 years continued uptime as a WHOIS server for a small domain registrar.

[aegersz]

OMG - 3 years ? what the hell am i worried about ? that's amazing, VERY amazing.

i thank you all for putting my ill founded reservations away.

... but even then, i always get nervous immediately after some admin changes when I issue the reboot command; my server is in Singapore, I am in Australia and most of my members in America ... never having even seen a picture of my server doesn't instill confidence in me but that is changing.

i am glad that i started this thread because my expectations were not realistic

[青山 素子]

OMG - 3 years ? what the hell am i worried about ? that's amazing, VERY amazing.

Terrifying, actually. Luckily it had a lot of security around it and was only providing WHOIS services. Data flow was also one-way to prevent a compromise on that server from affecting the actual registrar functions.

The more concerning server was a web server that was up for over two years straight without any kind of patching. Basic firewall, but no security beyond that. It made me very nervous. For that server, I made sure to get it in writing that there were known and unknown unpatched security issues in my request for maintenance. The clients never approved the work needed to actually update it.

Except for very specific circumstances, you want to make sure your system is properly patched at all times. This is especially important for public-facing systems. You'll normally be looking at 20-30 day uptime if you follow that practice.

... but even then, i always get nervous immediately after some admin changes when I issue the reboot command; my server is in Singapore, I am in Australia and most of my members in America ... never having even seen a picture of my server doesn't instill confidence in me but that is changing.

Sometimes you don't want to see a photo. Believe me on that.

The important thing is to have console access in case you manage to mess up the networking config or have to deal with a frozen system. That can be via built-in server hardware (Intel offers it on some of their higher-end boards) or an IP KVM.

Where I currently work, we have two very kick-ass servers (I think they're Dell PowerEdge M620 blade servers). I've never seen them, as they're in Texas and the company is in California. They run VMWare ESX and have over 7 VMs between the two with room for much growth. Hardware management is handled by the data center and I have "console" access to the VMs using vCenter. It works out pretty good for us, and it's been reliable (when the hosting company manages to not be idiots). More than enough control for what we need, and no need to see them physically.

[aegersz]

luckily I have managed to not screw anything up.

if I lose ssh access then I use webmin's command shell to drop the iptables, restart sshd etc.

my Host can force a reboot but I have no console access.

that's why I get nervous

... but you have prompted me to look into getting access.

thank you.

[aegersz]

> uptime
16:33:45 up 24 days, 10:03,  0 users,  load average: 0.24, 0.16, 0.10

still going strong !