Posts made by Tribble

Just updated to 1.1.1 due to other issues, I’ll retry on the newer build when i get a chance.

Hi guys,

I ran a 1.0.1 full registration w/ image task which failed. The issue that caused it had been fixed in a later release so i went ahead and updated to 1.1.1

Now my problem is as follows.

When i did the above registration task and it failed, it created a device in fog. That device cannot be removed, or edited (allows into edit but update never works, doesn’t even show the host name in there).

I tried entering the data for it manually, but it doesn’t work. Clicking update just takes it to a message that says “[FONT=Ubuntu][COLOR=#555555]No user history data found!” Clicking delete takes it to the remove page, where i hit remove and the following appears in the address bar, but nothing happens.[/COLOR][/FONT]

[FONT=Ubuntu][COLOR=#555555]http:\SERVERIP\fog/management/index.php?node=host&sub=delete&id=0[/COLOR][/FONT]

[FONT=Ubuntu][COLOR=#555555]Obviously, the registration task somehow created a device with no ID field in the database. Assuming that the page code uses that key field to perform the delete, it is unable to find the entry correctly.[/COLOR][/FONT]

[FONT=Ubuntu][COLOR=#555555]How can i get rid of this?[/COLOR][/FONT]

I’ve only been able to test it so far on 2 different models.

HP DC5750s = works
HP DC7900s = fails as described

Yes, i do get a download prompt.

It is set as memtest.bin

When trying to pull up Memtest from the 1.0.1 default boot menu the screen freezes at:

Loading boot sector… Booting…

Finally managed to get around to looking at those switches. Disabled Spanning Tree Protocol, and they now boot correctly.

Thanks Tom.

So, you’re saying that the issue would only appear when a PXE booting device using Undionly.kpxe is directly attached to a managed switch with STP enabled. And that adding an unmanaged, Non-STP switch as an intermediary between the PC and the managed switch can prevent the problem because iPXE communicates directly with the switch for DHCP? Therefore since the PC is not “directly” communicating with a switch that has STP enabled, there is less delay establishing a connection with the DHCP server.

Now, I will admit that i do notice a shorter time to pick up DHCP when i have the intermediary switch in place, but i never realized that switches had anything to do with DHCP packets and requests other than routing them to the correct ports.

I won’t be able to get on those switches until possibly friday due to our DR testing schedule, but i’ll be sure to let you know what happens. I know we haven’t done much if any customizing to those managed switches from their default state, so STP being enabled is certainly possible.

[quote=“Tom Elliott, post: 29104, member: 7271”]In reference to the “twirling” you’re experiencing, do your switches have STP enabled on them? Do you have an option for portfast on these switches?

Being connected to the 10/100 doesn’t really mean anything, it simply means it can receive the dhcp request from ipxe. If you take out all managed switches and connect directly to the FOG server, does all work as expected at 1gb?

This will let you know if it’s a 10/100->1000 problem. I run gig switches on all my tests, and don’t experience the boot loop issue. (BTW for all, the boot loop is intentional kind of).

Doing this won’t necessarily be easy, but I suppose, if you could for DHCP, connect your FOG Server, Client test system, and DHCP server to the same switch. Start off easy and use a “dummy” switch preferably rated for 1GB.

This will, again, determine if it’s truly the undionly OR if something’s on your network not sending the DHCP back in time.[/quote]

Hey Tom, I’ll see what i can do about running them all through a single unmanaged gigabit, but i doubt i’ll be able to do that without significant network disruption. Nearly everything runs through those 2 Dell powerconnects.

What i can do at the moment, is try an un-managed gigabit capable switch in place of the 10/100 C-net.
I’ll also have the Network admin check the managed switches for STP and portfast.

Keep in mind however this issue never came up while using .32.33 to image well over 100 pc’s this spring for our Win-7 roll out. We’ve had no changes to our network infrastructure since then. Those 100+ PC’s were loading through with no issues up until i updated FOG to 1.0.1 .

Appx 25 of the machines mentioned are in this office, i have not verified that all of the PC’s in this office are effected. I have not updated the DHCP PXE image name at our other 14 branch offices as of yet as that’s just way too many “OMG my computer won’t boot” phone calls every morning lol. Each office has it’s own subnet (10.X.0.x) and Local DHCP server.

All in all, i’ve imaged and installed 80 DC7900’s, and 80+ DC5750’s since january using fog .32 before the update with no issues like i’m describing here. A PXE\STP issue would have reared it’s ugly head already.

I’m certainly not complaining, Fog has already saved me hundreds of hours of work, lol, I just want to give you as much info as i can since i’ll be pretty busy today with DR testing.

When connected to the 10/100 switch everything functions as well as can be expected. They boot just fine right into the menu and then on to windows.

Not that it’s necessarily relevant yet, but it locks up at “loading boot sector… booting…” when trying to do a memtest. Let’s not trouble Tom with other issues here until we can get everyone to the menu first

However bump up the speed to 1gb, (by removing the 10/100 switch) and it, picks up DHCP, goes to configuring, twirls there for a bit, then flashes an error message which i can’t read, because it instantly reboots.

The netgear switches are isolated to our DMZ, i didn’t take the time to trace all the cables. But to allay your fears, .33 worked perfectly for all our machines w/pxe.

The “Production” desk is directly connected to a dell Powersafe, that i know for certain.

This is a recent issue since 1.0.1 and more specifically, if STP were an issue i would not expect it to suddenly “go away” just because i put another switch in-line after it. In fact i would expect it to get worse or at least stay the same since there would be more usage of STP with another switch in place.

In this particular office, we only have 2 subnets. Our DMZ, and the main network. In fact, i only ever use Fog here in this office to clone new machines, and i’ve done that well over 100 times on .33 with no problems

Well, other than multicast taking the network down when we try to use it, but that’s neither here nor there and probably has more to do with our highly distributed network and massive VPN usage to handle it all. In either case, i’m not worried about multi-casting, and we don’t push images over the WAN anyway.

Wow never expected that result.

Yes there appears to be something in the network itself causing this newer problem. Took my test machine out to the “Production” workstation and it too failed to boot (reboot loop). Brought it back to the “Test” desk and it works fine.

Grabbed the box from the “Production” workstation and brought it to my “test” desk. It booted up flawlessly.

Upon initial inspection, the only major difference in network protocol is that the “Test” desk is 10/100, while the “Production” desk is running on gigabit.

Further inspection denotes that all the core switches are either Dell Powerconnect, or Netgear Prosafe Gigabits. The “Test” desk has a cheap-o C-net 10/100 24 port to give me the number of ports i need at the desk which effectively drops the network speed to 10/100 for that cable run. The “Production” desk is running directly off the core switches at full Gigabit speed.

I then pulled the cable from my 10/100 switch and tried booting a pc directly on the “Test” desk from the direct line, without the 10/100 switch reducing the transmission speed. Reboot loop occured once again.

Therefore i must conclude that there is an issue with Undionly and gigabit speeds.

I think i may have spoken too soon about my HP 7900’s. It works fine on a box i’ve been building for a new user. But several of them in production had issues with Looping restarts this morning. Escaping out of PXE boot mode allows them to get going.

There are no hardware differences between the boxes. Only their physical location is different, and therefore they may not share some of the same switches/networking gear. They are both in the same office building, network, subnet, DHCP server, are all the same.

I’m going to pack the new box out to the station that was having trouble and test it there to see if that makes a difference. I’ll let you know.

This newest Undionly.kpxe is working for my DC 7900’s.

Hi tom,

I’m getting “could not start download, operation not supported” error as well after upgrading.

PC: HP DC 7900s w/ onboard Intel 82567LM-3
Fog Server: VMware, running Ubuntu 12.04LTS & Fog 1.0.1

Files tried: Stock, 1312, & IPXE.pxe

Apache error log

[CODE]
[Wed May 21 09:13:39 2014] [error] [client 10.99.0.75] PHP Warning: mysql_real_escape_string(): A link to the server could not be established in /var/www/fog/lib/db/MySQL.class.php on line 253
[Wed May 21 09:13:39 2014] [error] [client 10.99.0.75] PHP Warning: mysql_query() expects parameter 2 to be resource, boolean given in /var/www/fog/lib/db/MySQL.class.php on line 95
[Wed May 21 09:13:39 2014] [error] [client 10.50.0.52] PHP Warning: mysql_query() expects parameter 2 to be resource, boolean given in /var/www/fog/lib/db/MySQL.class.php on line 95
[Wed May 21 09:13:39 2014] [error] [client 10.50.0.52] PHP Warning: array_key_exists() expects parameter 2 to be array, null given in /var/www/fog/lib/db/MySQL.class.php on line 156
[Wed May 21 09:13:39 2014] [error] [client 10.50.0.52] PHP Warning: mysql_real_escape_string(): Can’t connect to local MySQL server through socket ‘/var/run/mysqld/mysqld.sock’ (2) in /var/www/fog/lib/db/MySQL.class.php on line 253
[/CODE]

Updated:

The following will now ignore the case of the Computer name by adding the /I argument to the If statement.

[INDENT=2]@Echo Off[/INDENT]
[INDENT=2] [/INDENT]
[INDENT=2]Set csv=\Server\Share\pcsetup\ipchanger\ComputerList.csv[/INDENT]
[INDENT=2] [/INDENT]
[INDENT=2]for /f “skip=1 delims=, tokens=1-7” %%a in (%csv%) do ([/INDENT]
[INDENT=2] if /I “%computername%”==“%%a” ([/INDENT]
[INDENT=2] [/INDENT]
[INDENT=2] netsh int ipv4 set address “Local Area Connection” static %%b %%c %%d[/INDENT]
[INDENT=2] netsh int ipv4 set dnsserver “Local Area Connection” static %%e[/INDENT]
[INDENT=2] netsh int ipv4 add dnsserver “Local Area Connection” %%f[/INDENT]
[INDENT=2] netsh int ipv4 add dnsserver “Local Area Connection” %%g[/INDENT]
[INDENT=2] [/INDENT]
[INDENT=2] )[/INDENT]
[INDENT=2])[/INDENT]
[INDENT=2]REM Pause[/INDENT]
[INDENT=2]Echo On[/INDENT]

Windows Vista and up have some built in VB scripts now to handle printer installations. Perhaps a rewrite can be done to use those built in scripts? Here’s a sample of a local install script using those.

[CODE]@ECHO OFF
md C:\DRIVERS
md C:\SCRIPTS
xcopy "\server\e\PrinterDrivers\HP Universal Print Driver\pcl6-x64-5.7.0.16448*." C:\DRIVERS*. /e /h /r /k /c /I /y /D
xcopy “\server\e\printersetup\scripts” C:\Scripts /e /h /r /k /c /I /y /D
cscript “C:\Scripts\Prnmngr.vbs” -d -p “PRINTER NAME”
Cscript “C:\Scripts\Prnport.vbs” -a -r IP_10.1.0.159 -h 10.1.0.159 -o raw -n 9100
Cscript “C:\Scripts\Prndrvr.vbs” -a -m “HP Universal Printing PCL 6 (v5.7.0)” -i C:\DRIVERS\hpcu155u.inf -h C:\DRIVERS -e “Windows x64”
Cscript “C:\Scripts\Prnmngr.vbs” -a -p “PRINTER NAME” -m “HP Universal Printing PCL 6 (v5.7.0)” -r IP_10.1.0.159
rem rd C:\SCRIPTS /s /q
rem rd C:\DRIVERS /s /q
@pause
[/CODE]

Now this is a .bat which Copies the required script files (needed for XP/2000) and printer drivers locally, then installs the printer locally.
Included is a .zip of the printer management scripts, but you can find them in the system32 folder of any version of windows vista and newer.

[url=“/_imported_xf_attachments/0/421_Printer.zip?:”]Printer.zip[/url]

Our images are 16GB Singlepart using Win7x64 with a full compliment of pre-installed software. I can image 7-10 computers at once in under 8 minutes without multicast. What in the world are you installing on those images to make them that big LMFAO

[quote=“aaronrus, post: 9276, member: 2756”]When adding a network printer in FOG 0.32 extra backslashes are added to the alias field. For example \\server\printer would be added to the database when \server\printer should be added. This is due to the line $alias = mysql_real_escape_string( $_POST[“alias”] ); in printer.add.include.php by changing this line to $alias = $_POST[“alias”]; The printer is added correctly \server\printer but doing this leaves you open to SQL injection attacks and would not be recommended.[/quote]

[url]http://fogproject.org/forum/threads/printer-manager-windows-7-service.395/[/url]

Looks like somewhat of the same issue. I’m getting this too with the ISDS. Injection isn’t a concern for us as Fog is an internal only server, I’ll give the rewrite a shot.

[url=“/_imported_xf_attachments/0/417_ISDS.JPG?:”]ISDS.JPG[/url]

I was hoping for an interface element to make modifying the .csv a little less confusing. A tool does make perfect sense for those IT departments whose managers rely heavily on Static IP’s like ours.

In my industry, reliability is paramount and we don’t take chances. If a branch server goes down and can’t assign an IP that’s a problem and we avoid potential problems like the plague.

Yes, i know you can set a static IP from DHCP, Try convincing the paranoid CTO of a bank that it will work just fine lmfao. he keeps giving me this deep dark glare like i’m insane for even considering it.

(it took me 8 years to convince him that we can reliably use Print servers to manage our printers instead of setting every single one by hand on every PC we installed *i ended up having to make scripts for that too finally until a month or 2 ago XD * and I only managed that miracle after i demonstrated how to use GPO to re-direct them all to another print server in under 10 minutes)