6 replies to this topic

[40H3X]

First: I know this topic is slightly "sub" related to PLi and hope it's not instantly removed by the moderators, but if so I'll understand

 

I had a discussion about the use of scheduled reboot and would like to know about the pros and contras or the sense or nonsense. To make a start...

 

PRO's

Scheduled reboot is often used to give the routers a little clearing now and then to keep its performance at its peak.

 

Contra

 I find it's really unnecessary though, there should never be any reason you need to reboot every day, or ever for that matter, unless you make substantial config changes that require it or a firmware update

 

[Erik Slagter]

Professional routers never require rebooting, that's a requirement a customer may expect to be met. Home-grade routers also don't require reboots, when no memory leaks are present in drivers. Home-grade routers are almost always made of an ARM SoC board with a few interfaces and a linux kernel. You may expect that linux doesn't loose memory, so if the proprietary interface drivers are well-written OR if they're using interfaces supported out of the box by the linux kernel, AND if the management application (webif etc.) doesn't leak memory, a reboot is never necessary.

 

Something similar goes for STB'es. If all drivers are written well, a reboot is never necessary. But in practise that's not always the case, unfortunately.

[MiLo]

I replaced a switch today and moved things around, otherwise my modem would have had an uptime of several months. I never reboot it for other reasons than power loss or so.

The receiver in the living room also never reboots without need. I don't update it much, so it serves as a stability test for E2. Typically I upgrade it about once a month, so on average it will have an uptime of several weeks. During some home renovation, that receiver was in the attic, connected to the dish, a router and power. I had no way of reaching it for almost two months, so if it needed a reboot, I would have to tear down the ceiling or so to get to it. It recorded everything neatly during the whole renovation, no problem whatsoever. It was quietly doing its work for over 3 months, serving recordings to the box with a wifi stick in the trailer outside.

My record for a Windows PC (running NT) is 1,5 years by the way. A normal desktop PC at work, which I never rebooted or shut down, and it was being used for development for 40 hours each week, and ran some background stuff for the remainder of the time. It had an uptime of 1,5 years when the IT department decreed its software should be updated and it was forced to reboot.

I agree with Erik, but I'd go further.

If it needs to reboot, it is broken.

Edited by MiLo, 18 January 2014 - 20:07.

[Erik Slagter]

I agree with Erik, but I'd go further.
If it needs to reboot, it is broken.

Actually that's exactly how I think about it.

 

I once maintained a server that had 20-40 users logged in and active at any time (with a specific monitoring application). It was running SunOS 4, imho the most stable OS SUN has ever produced. It ran for FOUR YEARS without interruption. At that point it had to be relocated, otherwise it might have run even longer.

 

Various linux servers at my current occupation have been running for multiple years, with records of 3.5 years, mostly because the server needed to have a new kernel.

 

The problems always begin with proprietary closed source drivers, that cannot be reviewed and corrected by the open source community. And unfortunately that is what we're stuck with, with our STB's, it makes me sad.

 

Oh btw, before I stepped over to cable, I was using adsl, using a adsl modem (not router!), and it ran for over a year without rebooting, as well. The (consumer grade) "internet router" we're using at my current occupation for offering wireless internet to customers, hasn't been rebooted as well, it has been running for almost a year now.

Edited by Erik Slagter, 18 January 2014 - 20:27.

[WanWizard]

For me a kernel update is the only reason for a restart. Since most my systems are (in one way or another) internet facing, I do that quite regularly, I can't afford not doing that.

 

Otherwise, years of uptime is not an issue. Like the server that runs this forum's database:

 

[root@database3 ~]# uptime
 00:24:29 up 503 days,  6:24,  1 user,  load average: 0.34, 0.18, 0.12
 

[Erik Slagter]

I've been looking around, and we have a winner! This machine will be switched off soon, that is going to hurt. Despite the low load at the moment, it's quite a heavily loaded machine...

 

09:48:40 up 1202 days, 22:11,  2 users,  load average: 0.03, 0.02, 0.00

Edited by WanWizard, 19 January 2014 - 10:20.
translated, it's English here...

[40H3X]

Thanks for everyone's input it may (or may not) come to no surprise that my opinion was the Contra and I agree with everyone (thus far) that there are no reasons for periodic reboot, other than faulty hardware/firmware. Yet this advise is given widely on CS sites with the argument that it "clears things"??? TMHO the use of "clearing things" this way is the result of flaky programmers often encouraged  or better pressured by management by setting a too narrow timeline for development, but hé that's the power of money and therefor I would go so far as that the presence of this feature is already ill news to what can be expected from the device.

 

Yet, I hoped that there would be some Pro's, but then again it's all hobby here so there is no pressure from management only sometimes nagging users