[CFT] Need Testers for: sysutils/bsdconfig

[Devin Teske]

On Jun 22, 2012, at 12:07 AM, Eugene Grosbein wrote:

> 22.06.2012 10:50, Ron McDowell пишет:
> 
>>> Again, thank you very much for testing this new software.
>>> P.S. Due to the large codebase comprising bsdconfig, ample precautions should be taken. I've not noticed any negative behavior in months of usage, but just be warned.
>>> 
>>> P.P.S. I don't think on subscribed to -stable@, so include me in your replies.
>> I'm one of the coauthors of this code, and I am here on -stable.  As 
>> stated, this port will only run on 9.0-RELEASE and later.
>> 
>> Please give it a try!  Thanks!
> 
> I've tried it on installed 9.0-STABLE system.

First, thank you for testing!


> 1. At the moment, the documentation does not install:
> 
> Could not install package en-freebsd-doc (pkg_add: unable to fetch
> 'ftp://ftp.freebsd.org/pub/FreeBSD/ports/amd64/packages-9-stable/Latest/en-freebsd-doc.tbz'
> by URL)
> 

Ok. Thx.

A PR should be filed against bsdinstall (since bsdconfig is simply linking to "bsdinstall docsinstall" for that menu item and is not actually responsible for the bug).


> 2. For system running single gmirror gm0 consisting of 2 whole disks
> ada0 and ada1, 'Label allocated disk partitions' offers
> only ada0 and ada1 for disklabel editor and not gm0.
> And disklabel editor shows empty list of partitions.
> 

Ok, Thx.

This too is not part of bsdconfig, but is linked to sade(8).

NOTE: "Documentation Installation" and "Disk Partition/Label" are the only two modules that are pointed to other tools (the former linked to "bsdinstall docsinstall" and the latter linked to "sade").

The goal for this is to replace sade (which is not geom compatible) with a new tool.

This is on the community's to-do list.


> 3. Similar for FDISK Partition Editor (ada0/ada1 and not gm0 for choise),
> but Partition Editor presents whole disk space as 'unused'
> 
> Note that current SATA disk prices and quality offers no choice for servers
> other that some kind of RAID.
> 

Again, this is because sade(8) is not geom compatible.


> 4. Networking Devices configuration presents lagg0 device as <unknown network interface type>.
> It should provide descriptive string like 'Link aggregation/failover'.
> 

It's on my to-do list to change the way descriptions are calculated.

Currently, I've got a static hard-coded list of descriptions. Someone recommended that there was a way to get this information from sysctl.


> 5. Same for vlan16. For vlan9 is shows right 'IEEE 802.1Q VLAN network interface'.
> It should work same way for vlan1-vlan4095 interfaces at least.
> 

I'd like to know if the sysctl MIB's for describing network interfaces is reliable. Maybe I'll keep the static list as a fallback. But yes, you're absolutely right -- I should have supported up to 5 digits even (ifconfig has internal limits of 16-bit unsigned integer for the interface instance-number).


> 6. Same for ipfw0 pseudo-interface.
> 

Curious what sysctl says about it.


> 7. Networking Devices configuration does not allow to configure any interface
> while there are mounted NFS volumes. Should present a warning only, not disallow the operation.

Did I completely disallow it? I'll have to re-check -- I thought that I had made it so that you could view/edit the configuration but that the warning says that changes will not become effective until you either reboot or visit the menu again when no NFS mounts are active.


> For example, it should be possible to configure new vlan interface while NFS mount
> uses another clan.
> 

Do you know of a handy way of determining which NFS mount is using which network interface? And further, is there a handy way of traversing the route path to determine that one interface isn't required as an intermediary transit device? (meaning: can one truly ever know that making a new configuration active on any interface could not potentially drop your entire machine from the net with hung NFS mounts?)

Many months of testing in the lab produced no less than 6 edge-cases where -- if a network link or route is modified when NFS mounts are active -- the machine can enter an unstable/unusable state.

So we decided to err on the side of caution when it came to allowing settings to be made-active when NFS mounts are active.

I'm not against improving the code, but I'm wondering if it wouldn't be safer to stick to disallowing any/all changes from being made-active (while allowing viewing/editing without making-active) when NFS mounts are active.

NOTE: There are other safe-guards too. For example, if you're logged in via SSH and using X11 forwarding while passing the "-X" flag (to use Xdialog(1)), you are disallowed from making a new hostname active (you can change the hostname, but not make it active) because that would cause the very next iteration of Xdialog(1) to fail due to a surreptitious X authority revocation based on the hostname-change in mid-session.


> 8. In DNS Nameserver Configuration, it's not clear that one, in fact,
> can remove unneeded DNS server through two-step procedure - first try to edit,
> then clear the address. It should be more obvious at first.
> 

Can you have a look at "bsdconfig startup_rcconf" and see if that's a better way to go about the deletion-process?

Or perhaps you're just advocating a helpful message in the text above the menu list that explains how to delete the item? (least amount of work)


> That's enough for this time :-)
> 

Thank you again for testing!
-- 
Cheers,
Devin

_____________
The information contained in this message is proprietary and/or confidential. If you are not the intended recipient, please: (i) delete the message and all copies; (ii) do not disclose, distribute or use the message in any manner; and (iii) notify the sender immediately. In addition, please be aware that any message addressed to our domain is subject to archiving and review by persons other than the intended recipient. Thank you.