New shell server transition

[roelofs]

scott wrote:

• Currently, there isn't an auto-logout set by default. (You can set your own with tcsh or bash, though.)

Do we want there to be an auto-logout, and if so, what do folks think the timeout value should be?

There is an auto-logout, and it's driving me nuts. oldshell doesn't have one, but not even a vim or tail session on sh disables its auto-logout. (Either one suffices on Linux hosts at work.)

I don't care if sh defaults to auto-logout as long as I have a way to defeat it. I really, really do not want to have to re-login multiple times per day, or even per month. :-(

Thanks,
Greg

Hi Greg,

I can't think of a reason why sh is hanging up on you. I will see if I can't track that down. It might have something to do with tcsh, since my bash sessions don't time out...

-Scott

[roelofs]

scott wrote:

• Currently, there isn't an auto-logout set by default. (You can set your own with tcsh or bash, though.)

Do we want there to be an auto-logout, and if so, what do folks think the timeout value should be?

There is an auto-logout, and it's driving me nuts. oldshell doesn't have one, but not even a vim or tail session on sh disables its auto-logout. (Either one suffices on Linux hosts at work.)

I don't care if sh defaults to auto-logout as long as I have a way to defeat it. I really, really do not want to have to re-login multiple times per day, or even per month. :-(

Thanks,
Greg

Hi Greg,

I can't think of a reason why sh is hanging up on you. I will see if I can't track that down. It might have something to do with tcsh, since my bash sessions don't time out...

-Scott

Hi Scott,

First of all, thank you. I stupidly assumed everyone had encountered this but wasn't as bothered by it as me...

I saw slightly different behavior in the most recent occurrence: the session was still seemingly active in my xterm, but the underlying SSH session had clearly died (keyboard unresponsive), and my local ssh client hadn't yet noticed the lost connection. That suggests it might simply be different sshd behavior w.r.t. inactivity, due either to newer code or to a change in the server-side config.

I vaguely recall an SSH client-side "keep alive" config or command-line option of some sort; let me go check with our googly overlord...

Ah, yes, I can add "ServerAliveInterval 120" (for example) to my ~/.ssh/config file to send a null packet to the server every 120 seconds, which supposedly will prevent such issues. (There's a corresponding server-side pair that can be set in /etc/ssh/sshd_config to do the same for all clients: ClientAliveInterval and, if a maximum number of these is desirable, ClientAliveCountMax. For example, setting those to 120 and 720, respectively, would keep clients alive for up to 24 hours before letting them die.)

Anyway, let me try that and see if it works...

Thanks again!
Greg

[bfmills]

I have a Wordpress installation. Will you be migrating our data from one server to the other? Will the public facing URL remain the same? Thanks.

[scott]

I have a Wordpress installation. Will you be migrating our data from one server to the other? Will the public facing URL remain the same? Thanks.

Hello,

The email I sent out was only to accounts where someone had actually called support and requested access to the shell server. I guess a lot of account holders have this access, but someone else on the account (say, a webmaster) had requested the access.

So I apologize for causing confusion. This won't affect web hosting, unless one's webmaster has created a crontab to run commands periodically. I just checked your account, bfmills, and your account doesn't have a crontab, so you should be good to go.

-Scott

[scott]

shellmaster@sonic.net now goes to our ops ticket queue, so you'll have other system administrators answering questions besides me.

-Scott

[roelofs]

There is an auto-logout, and it's driving me nuts. oldshell doesn't have one, but not even a vim or tail session on sh disables its auto-logout. (Either one suffices on Linux hosts at work.)

I don't care if sh defaults to auto-logout as long as I have a way to defeat it. I really, really do not want to have to re-login multiple times per day, or even per month.

Thanks,
Greg

Hi Greg,

I can't think of a reason why sh is hanging up on you. I will see if I can't track that down. It might have something to do with tcsh, since my bash sessions don't time out...

-Scott

[...]

I vaguely recall an SSH client-side "keep alive" config or command-line option of some sort; let me go check with our googly overlord...

Ah, yes, I can add "ServerAliveInterval 120" (for example) to my ~/.ssh/config file to send a null packet to the server every 120 seconds, which supposedly will prevent such issues. (There's a corresponding server-side pair that can be set in /etc/ssh/sshd_config to do the same for all clients: ClientAliveInterval and, if a maximum number of these is desirable, ClientAliveCountMax. For example, setting those to 120 and 720, respectively, would keep clients alive for up to 24 hours before letting them die.)

Anyway, let me try that and see if it works...

Works like a charm! I'll experiment with longer intervals, but two minutes is definitely sufficient. Happiness achieved.

Greg

[scott]

Got really busy tonight, still a lot of folks logged in:

Code: Select all

_[/etc]_(root@sh)_
# uptime
 01:08:46 up 15 days,  4:14, 32 users,  load average: 0.44, 0.71, 0.69

I'm thinking a 1-hour shell idle timeout might not be a bad idea. What do you think?

-Scott

[casner]

I'm thinking a 1-hour shell idle timeout might not be a bad idea. What do you think?

-Scott

Please don't put in an idle timeout that users cannot individually reconfigure to remove.

An aside regarding what appeared as auto-logout for Greg: If the connection was completely idle so there were no packets exchanged at all, then the connection may have dropped because the NAT function in his home router timed out the connection state.

[patty1]

Hi, Scott. A one-hour timeout sounds good to me. Although I probably won't need to customize that, I support casner's suggestion of users being able to change the timeout.

[scott]

Hi Scott,

Can we get ncurses installed please? Had it on the old shell, and I use(d?) it fairly often

Dan

It's already installed -- what problems are you seeing?

Code: Select all

Package ncurses-devel-5.9-14.20130511.el7_4.x86_64 already installed and latest version

-Scott