Recent Journal Comments

I modified the last two. On 4 clicks goes till the end of line with no white spaces. On 5 clicks; it selects the entire page; including scrolled lines.

xterm*on2Clicks: regex [^ \n]+ 
xterm*on3Clicks: line 
xterm*on4Clicks: regex [^*\n]+
xterm*on5Clicks:all

PS: Sorry; I am several years late for a reply. Next stop; back to the future. :-)
(go to entry which this is a comment to)

For whatever its worth, I was absolutely unable to maintain a connection to the wireless network at PyCon. I could get it to work for maybe 5-10 minutes then poof my laptop radio shutdown and required me to reboot my computer to get it to work.

The extricom stuff did something to my machine to cause all these problems. I have an Intel 3945ABG chipset with the latest driver software from Intel running on an HP laptop with Windows XP SP2.

Wireless of course worked fine in my room using the ibahn network provided by the hotel.

Just thought you should know...
(go to entry which this is a comment to)

The mouse pointer issue isn't really a big deal for me, because we rarely use the mouse with our Linux systems. We can do the installs in text mode, and when we can't the installers have extremely good keyboard acceleration so the installs can be done without a mouse.

We connect these only to servers, where we don't want a GUI, and our primary use is only when something is wrong, so it's very rare, and we probably want the text console anyway, not a GUI. So for this it's perfect.

You can try disabling the "pointer sync" or enabling it to see if that makes it better, but the problem as I understand it really has to do with how the PS2 standard for mice works, it's all relative rather than absolute, and with acceleration there's very little that you can do to make the mouse stay in sync with a VNC session.

Sean
(go to entry which this is a comment to)

Very informative review. I thought that StarTech would have solved the mouse pointer synchronization issues by now. After upgrading the the latest "firmware" the problem remains. I'm waiting for my second email from StarTech support after reporting the persistent pointer sync issue.

I thought it strange that the "most recent" firmware release is still over 1 year old.

Component		Version / Release
System firmware		Wed Mar 7 14:40:13 EST 2007
CGI Component		07.10.3143245
Linux Kernel		Linux version 2.4.20-pre7 #144 Mon Oct 24 16:07:46 EDT 2005
System FPGA		17     
Model name		SV1110IPEXT (startech.com)
Software options	0000001f (ENT, SEC, MULTI, IPMI, MODEM)

Has there been any progress on this pointer sync issue? How does anyone connecting to a windowing environment get around this?! Don't people ask for their money back? This is a real show stopper if one cannot use the mouse!

Thanks for the great review. I'm interested to know if you've heard anything else on this, though, like under what conditions does this actually work?
(go to entry which this is a comment to)

I appreciate the ssh hack you published as well as the superb review. It gave me some insight into how they've put together the system. I'm interested in getting some additional functionality out of the 3 SV1110IPEXT units we've purchased for use at a client site.

My situation is sort of backwards from how most users employ the StarView. I need multiple users to be able to access a single (Fedora Linux) machine for Internet access, when all they can see on the stand-alone, private network is the StarView's SSH port. I'd like to tunnel a VNC port through the StarView directly to the Fedora Internet machine. I'm using the Xvnc module for the X server to allow multiple VNC sessions.

The Fedora machine has two NIC's one, attached to a firewall and the other directly connected to the StarView WAN port. The Keyboard, Video, and Mouse are all attached to the StarView. Only SSH traffic may pass between private clients and the StarView. Only SSH traffic may pass between the StarView and Fedora Internet machine.

Can you suggest a method to configure or modify the StarView machine to allow routing SSH traffic between the LAN and WAN interfaces?

Ntwk1: Private, stand-alone network: no outside network connectivity Ntwk2: Private, network: firewalled Internet access

1. Connect the LAN port to Ntwk1
2. Connect the WAN port to the 2nd NIC on machine in Ntwk2
	1st NIC goes to switch on Ntwk2 or directly to private firewall interface

Port 5973 is setup to in xinetd to accept VNC connections for 1280x1024 res, 24-bit color depth

[Ntwk1|Switch]->[LAN:StarTechIPKVM:WAN]->[Fedora WS|Ntwk2]

[Ntwk1 client:22(localhost:15973)]->[LAN:22|WAN:22>]->[22:(fedora:5973)svr|Ntwk2]

##!! Important: No data transfers between Ntwk1 and Ntwk2 - only VNC over SSH

Firewall on Fedora WS permits only SSH and VNC to pass on 2nd NIC.

No routing at all between two NIC's on Fedora WS. IP KVM only accepts SSH & https on WAN and only from a single IP
(go to entry which this is a comment to)

I've put the slides up at under the "Free Stuff" portion of our web site at http://www.tummy.com/Community/Presentations/sysadminpython-200803/index.html, thanks for the reminder.

Sean
(go to entry which this is a comment to)

Sean,

Will you be posting your presentation somewhere?
(go to entry which this is a comment to)

Another strange thing about ipmilan's behavior is that stonith with -T on turns the system off, and with -T off turns the systems on... It seg faults in all cases, but only after the good deed is done, so for the time being, I'll live with it.
(go to entry which this is a comment to)

Thanks Sean,
I was just about to do an ltrace myself after having no success with strace. Anyway, using numeric values instead of symbolic names for auth and priv, seems to have solved my problem!
(go to entry which this is a comment to)

It seems fine. It's not great, but I wouldn't call it particularly slow either. Though I haven't really measured it, I just use it. I do know that in the "safe" mode it is very slow. I set it for "balanced" and it seems fine with a BBU. I would agree that it's not up to the performance of the marketing claims 3ware makes, but it's a rock solid product under Linux and that's worth some performance to me.

Sean
(go to entry which this is a comment to)

Are you happy with the performances on random I/O? Writes? iowaits?

I had a 9550 and was very unhappy (140 random IOPS on a 6-drives RAID5!), just traded it for a 9650 which is now under test (see http://www.makarevitch.org/rant/3ware/ )
(go to entry which this is a comment to)

Excellent work :) Saved me a lot o heartache!

FYI this is pretty robust, the my DC noc noticed my server down, and rebooted it midway through the process, upon reboot they noticed my rather obvious comments (added to init) spewing across the screen about resizing in progress, and contacted me to see what was going on! The process completed no problem :)

Do you happen to have a patch for resize2fs? I'd like to make a few binaries for various systems I have which have ridiculously large / filesystems.

Cheers!
(go to entry which this is a comment to)

There is a lot of interesting stuff. And if you make a list for yourself that is a little verbose about what work is there to do and why it would be cool to do it, I might like to see that and try to re-use it as a "cool crap for anyone new to do list" and start challenging people to get involved.

But whatever you do, do not overcommit. Please make an effort to set some time limits on how much your involvement takes and plan to stick to it. It's real easy for passionate people who care about a project or organization to feel like they need to fill in all the gaps simply because they see the gaps and they care so much.

Don't do it, don't fall into the trap of over-commitment. Figure out how much time and energy you really have to spend before you decide where to spend it. We will always have more work than people to do it. What we desperately need to make sure of is that the people we have who are doing the work are happy and excited.. instead of burnt out.
(go to entry which this is a comment to)

very good idea, but on kernel 2.4 resize2fs doesn't work

[root@linux root]# ./resize2fs
FATAL: kernel too old
Segmentation fault

(go to entry which this is a comment to)

I've also found digikam to be great for tagging etc. 0.9.2 had speed issues with large albums but they're fixed in 0.9.3 It would help if there was a faster keyboard interface for tagging lots of photos but its tolerable Also the tags are put into the photos using IPMC, and there extensions to gallery2 that support reading tags out of IPMC on upload - so I can tag in digikam, use the kipi plugin from digikam to export to gallery2, and all the images on gallery2 are automatically tagged and searchable... If I actually get round to uploading :-)
(go to entry which this is a comment to)

You can try digikam, it has tags, sorting by date, fullscreen diaporama, ... and is in active development. That's the best I tried.
(go to entry which this is a comment to)

You may have figured this out already, but I will post it for the benefit of your users.

If you run Stable, and that works well for you, but you encounter an issue where you need newer packages while trying to install something, you have the option of using apt pinning to select specific packages from testing or unstable. It is actually one of the most useful things I've ever used, but if not careful you can break things. So users should know what they're doing before attempting it.

For further information, readers can visit: http://jaqque.sbih.org/kplug/apt-pinning.html

I have found that to be an excellent beginners tutorial on pinning.

Cheers.
(go to entry which this is a comment to)

The problem is that the "phone company" does not want to turn into just another data pipe. They are convinced that they must also control your content, hence all the walled gardens. Maybe the Open Handset alliance can break this trend.
(go to entry which this is a comment to)

There also seems to be an open-source "megactl" (with fewer features): http://sourceforge.net/projects/megactl

And, it doesn't hurt to vote: http://www.ideastorm.com/article/show/68090/Leverage_LSI_to_Open_Source_MegaCli
(go to entry which this is a comment to)

Thanks! Your message saved me a lot of time. It's a mystery why Dell does not have a download page for MegaCli, or at least an information page.
(go to entry which this is a comment to)

This worked great..! I was a little worried to run this on my server that's far far away in another land, but it worked perfectly.

I originally tried this on my Fedora 8 box, but it was saying invalid partition size. It worked fine on my CentOS5 server.

Thanks!
(go to entry which this is a comment to)

The aforementioned bug 124789 has a recently consolidated patch to support both encrypted PVs (as you've done) and encrypted LVs. It's worth explicitly noting that an encrypted PV allows suspend and resume from a swap LV in the PV, while still only requiring one passphrase to unlock the whole system; the current Fedora encrypted swap partition is randomised on boot which borks suspend.

You can also use LVM tools - pvmove - to transfer the system from the unencrypted partition to the encrypted PV, as described in this HOWTO (this may have been the HOWTO you reference, though there are a couple attached to the bug).

I too am surprised this isn't being pushed as an install option, especially for laptops. There seems to be quite some resistance to including this feature in Fedora.
(go to entry which this is a comment to)

What about a tag cloud idea.... As the task ages it Gets a larger font like a tag cloud the larger font tasks are ones that you have to deal with. You could even change the color from green to red as the task become more important. I could envision a web page where you drill down to see tasks when you click on the title in your tab cloud.
(go to entry which this is a comment to)

Note that rpm has a poorly mentioned and documented option here too...

--last Orders the package listing by install time such that the latest packages are at the top.

ie, 'rpm -qa --last | less'
(go to entry which this is a comment to)