Simplysimple.info » Troubleshooting

Recovering from a broken array : Ubuntu Software RAID

Benjy — Mon, 17 Jan 2011 09:44:34 +0000

I have been using software RAID ( level 1) on our cheaper IBM tower servers which did not come with a RAID controller. This time when my ancient laptop failed for the umpteenth time, I gave up and decided to use one of our redundant servers which once ran Debian’s E-box as my Desktop. This model is an IBM xSeries 206m tower with two 160Gb SATA drives. Since I’d be doing a lot of experimenting with various server technologies, I chose to install Ubuntu 10.04 LTS Server on it. Since it needed to double up as my personal machine too, I did a sudo apt-get install ubuntu-desktop to get a GUI onto it. A couple of distribution updates later I had my desktop showing up as Ubuntu 11.04 ,Natty Narwhal. However doing a cat /etc/lsb-release shows…

DISTRIB_ID=Ubuntu
DISTRIB_RELEASE=10.10
DISTRIB_CODENAME=maverick
DISTRIB_DESCRIPTION=”Ubuntu 10.10″

…So I’m not sure which version of Ubuntu this post should be attributed to, anyway, when I came in this morning, my system was dead with one of the IBM hard disks making a clicking sound. I expected as usual to be presented with something like “your RAID array has been degraded do you want to continue? y/n” but somehow, I was shown a blank screen -sometimes alternating with the Ubuntu GUI loader in horrible colors. By the time I was shown a y/n prompt, it had already timed out, defaulted to no and dropped me to a busy box. After a few reboots with the same result, I rebooted my “server-desktop” with the server installation disk in the drive.

I booted to recovery mode from the Ubuntu Server CD and got to a shell and by running cat /proc/mdstat I figured out that out of my sda and sdb hard disks the sda disk had failed as it was missing from the list. So here is what you need to do to recover from a disk failure in one of your RAID arrays or rather what worked for me.

Run mdadm –remove /dev/md1 /dev/sda [Replace the values md1 and sda with that of your broken array and damaged disk ]

Repeat this for any other array that uses the damaged disk (sda here) and then shut down the server

Replace the damaged drive with an identical spare, boot to recovery mode using the server CD again and use fdisk to list the partitions on the disk containing your data.

fdisk /dev/sdb [Replace sdb with that of your old hard disk -the one containing your mirrored data]

p (to dump the partition table)

Disk /dev/sdb: 160.0 GB, 160041885696 bytes
255 heads, 63 sectors/track, 19457 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x7a4d7a4d

Device    Boot      Start         End         Blocks        Id System
/dev/sdb1              1                 973       7815591      fd Linux raid autodetect
/dev/sdb2    *         974       19457   148472730    fd Linux raid autodetect

Note down the information corresponding to the information shown above in red. You will need it for the next step.

Next do a fdisk /dev/sda [Replace sda here with your new hard disk]

Create new partitions on the new hard disk by pressing n at the fdisk prompt. Note down the partition numbers as you create them and use the same start/end blocks (which you noted down earlier) as that of your working older hard disk.

Change the partition id by pressing t and type fd so that the partition type matches that of your old disk. Next press a to toggle the bootable flag of your partition that needs to be made bootable and give it the appropriate partition number (which you noted down when creating partitions).

Press w to write out the partition table and exit fdisk.

Next do a fdisk /dev/sda [Replace sda with that of your new hard disk] again and p to verify that the settings you put have been committed by fdisk and match that of your older working hard disk.You should find a * after the partition you set bootable.

Now we can add the new disk to the array with the command mdadm –add /dev/md1 /dev/sda2 [Replace md1 and sda2 with the values of your previously broken array and damaged disk partition]

Do this for any other RAID partitions that you have which were using the damaged disk and reboot.

If your server does not start up, your bootloader was probably located on the MBR of the disk that you replaced. If this is the case, boot again through the recovery disk. At some point you will be asked whether you want to activate the existing RAID configuration. On choosing yes, you should see sustained hard disk activity, which indicates that the mirroring has started. From the recovery options, choose to rewrite the GRUB bootloader onto the MBR of your first hard disk. Rebooting now without the recovery disk should get your server up.

Typing cat /proc/mdstat at a terminal should show you the progress of the mirroring process.

benjy@rabid-rover:~$ cat /proc/mdstat
Personalities : [raid1] [linear] [multipath] [raid0] [raid6] [raid5] [raid4] [raid10]
md2 : active raid1 sdb1[1]
7815488 blocks [2/1] [_U]

md1 : active raid1 sdb2[1] sda2[2]
148472640 blocks [2/1] [_U]
[=========>...........] recovery = 45.6% (67775488/148472640) finish=35.4min speed=37981K/sec

unused devices:

This is what worked for me. Hope it works for you too. I was glad to see my data and installed applications once again!

Note : Once the mirroring is complete, the _U showing in the line above the progress will change to U. In my case I’ve left the swap partitions (which was using a RAID array too) degraded as I plan to activate them as discrete swap spaces with the same priority to obtain a high performance swap partition. There is no point mirroring the swap as there is no useful data in it from a users point of view.

Taking screenshots with Lightscreen

Benjy — Tue, 28 Sep 2010 08:56:13 +0000

Screenshots are an important part of IT support. Whether it be for documentation from the IT persons end or the screenshot which an end user sends to the service technician which eases understanding of the problem or the error message as the scope for misunderstanding is reduced to zero. Taking screenshots however is not as simple as it seems.

Run Lightscreen, Click on Options to configure and hide to minimize this window.

In my IT career I’ve seen them all. Its not just the end users who submit unproductive screenshots, I’ve had my share of receiving screenshots from fellow IT managers which filled up my mailbox as each screenshot was around 5MB each. This was true especially when Windows had no screenshot tool and people pressed the appropriate key to take a screenshot and then pasted it into Microsoft Paint which saved it as a massive .bmp file by default. Some of the others sent it pasted in a Word document which left me squinting as there was no reliable way of enlarging the embedded image to view the error message.

The General Tab, where you specify the folder to save your screenshots. It is best to leave these at their default values

On Linux it has been easy to take a screenshot. When you press either the printscreen key to take a screenshot of your full desktop, or the Alt + Printscreen key for taking the screenshot of an active window, the appropriate screenshot tool pops up by itself so all you need to do is save it in the format of your choice.

For Windows, even though Vista and 7 have the Snipping Tool for selecting an area of the screen to copy, in spite of it having erasing and highlighting features, I prefer using an Opensource tool called Lightscreen. Also the Home editions of Windows do not seem to come with the snipping tool. Lightscreen comes either packaged in a Windows installer or in the form of a stand alone zip file which can be run from any media without installation or the requiring of administrative priviliges on the system.

You can download lightscreen from here. Once you have downloaded the appropriate version install and run it.

The Hot Keys Tab, Used to activate the keyboard shortcuts used to communicate with Lightscreen

Once it is running, you should see a screen like the first screenshot. We need to configure Lightscreen once though, so click the options button.

On the General Tab on the Options screen, Leave the default settings as they are. In case you want Lightscreen to startup with Windows, then you can select the checkbox Run Lightscreen at system startup.

On the Hotkeys Tab, I usually selct all the check boxes. The text next to the checkboxes is self explanatory. Remember that the “Print” stands for the Print Screen button on the top right hand side of your keyboard.

On the Options and the Advanced tabs, you can select the options that you are comfortable with. I usually do a lot of documentation, so I choose Include the mouse cursor in the screenshot in the advanced settings

Thats it. Click on the hide button and you are good to go. Pressing the appropriate keys will now save the screenshot in your selected folder. No more annoying save prompts. All you now need to do is email them to the concerned person.

Here is what happens when you press the printscreen key - The whole of the screen is captured...

The same screenshot with Alt + Printscreen. Note that only the active window was captured!

Vista Graphic drivers for the Intel D845EPI/D845 GVSR Motherboard

Benjy — Mon, 13 Sep 2010 06:05:49 +0000

Hardware and Software vendors push you to keep updating your hardware and software. Time and again business users, who have no interest in playing this catch up game, but in maximizing their investment get caught up in this net. Non profit organizations fall into this same category. When we upgraded our last remaining pentium 4 computer from Windows XP to Vista Business, I didn’t know that I was in a for a nasty VGA surprise. ..When I went to Intel’s site, the drivers were available only for XP. I sincerely hoped that Vista would figure out what to do and have a compatible generic VGA driver which would give me at least an 800 x 600 resolution with a decent number of colors. This apparently did not happen. It would either display in 640 x 480 at 16 colors or at 800 x 600 which was even worse that 16 colors and not even the borders of the buttons were visible anymore, so operating the computer was more of click and miss. I was left with 5 options…

Downgrade Windows to Windows XP
Buy a new motherboard or Desktop
Try and find a compatible driver
Buy a PCI or AGP card for the board
Switch to Ubuntu Linux

The first was not very feasible as that meant spending a lot of money + tax on a new copy of Windows XP. (It isn’t very comforting to know that we need to buy a FPP -read as more expensive)

The second was the last “fail-safe” option -plan B. The third option was the only one left out as the accounting package that needed to be run would not run on Linux and finding a compatible graphics card would be an equally daunting task. Step 3 is documented here in the hope that some other user not wanting to upgrade this Intel board could use them. So here we go!

Right click “My Computer“, click on Properties and then “Device Manager”

Right click your "Standard VGA" device under Display adapters and click on properties. Select the Driver Tab and click on Update driver.

Select, "Browse my computer for Driver software"

Select, "Let me pick from alist of Device Drivers on my computer"

Un-check "Show compatible Hardware" and Click on Intel Corporation in the Left Pane and the "82852/82855 GM/GME Graphics controller " Driver in the right pane and click Next

Finish the Driver install and restart your computer

Your Display Adapter name should now have changed and it should now show you the new driver you installed

[Solved] JRE persistent printing prompt

Benjy — Mon, 09 Aug 2010 09:46:35 +0000

Back to HIS Installation Home

We’ve had this annoying problem when printing from an applet from our web based HIS software. The applet security window pops up (we’re using Firefox) with the prompt, “The applet has requested access to the printer. Do you want to allow this action?“ It also has a check box for “always allowing this applet to access the printer” . This however is never remembered and the same prompt is presented every time the applet runs which takes either two clicks of the mouse or two keystrokes (spacebar and enter) to dismiss. This is highly annoying as it is used for displaying/printing the bill which happens roughly once or twice every minute. The right way to go would be to sign the applet. Here is a link to Sun’s (Now Oracle) online resource. As usual our developers took a shortcut. This might however help others who have similar problems but cannot sign the applet for various reasons.

For Windows:

The annoying java security window.

Change over to the security folder of your JRE installtion -default is C:\Program Files\Java\jre6\lib\security

Next, open the file java.policy for editing in Wordpad (Do not use Notepad). Scroll to the end of the file.

This is how the end of the file should look after modificaton.

Press enter after the semicolon on the line just above the }; and properly indent and type in the following line…

permission java.lang.RuntimePermission “queuePrintJob”;
Restart your computer and you are done! This was tested on Windows XP Profession SP3.

For Linux:

The java.policy file in Ubuntu 10.04 is found in the /etc/java-6-sun/security folder. The same change as for Windows (Shown in the above screenshot) needs to be done.

In a Terminal window, type the following command (The Terminal is found in the Applications > Accessories > Terminal menu)

sudo nano /etc/java-6-sun/security/java.policy

Insert the line, permission java.lang.RuntimePermission “queuePrintJob”; just above the closing }; (Refer to the above screenshot)

Press Ctrl + X and save and exit

Moving your WordPress blog

Benjy — Fri, 06 Aug 2010 06:02:11 +0000

Ever have your WordPress blog stop working with a database connectivity problem when you shifted hosting or changed the IP address of your server? This has happened to me numerous times -especially when using WordPress as a platform for an Intranet. Changing the IP address of the server for any reason has devastating consequences. This post deals with self hosted instances of WordPress on your own Linux server or on that of your web host. I have used MySQL query browser to access the WordPress database. You can also do this through PHPmyadmin or any other database editing tool. If you are proficient with MySQL commands, you can do this instead from the MYSQL command prompt on the server console or through SSH. In addition to this you need FTP or SSH access to your server for editing the WordPress configuration file wp-config.php.

Edit the 'wp-config.php' in the root of your WordPress installation directory. In this screenshot I am using Winscp (from a Windows machine). You can also use SSH (Putty from Windows) or FTP for the same.

As shown in the above screenshot, replace your old host name with the new one. In my case, the host name is an IP address. Remember that this host name is not necessarily the link to your website -unless both your database and web server are running on the same machine. This hostname is the name/IP address of your database (MySQL ) server. So you should change this only if you’ve changed your database server as well. In my case, everything is running on a single server, so when I change its IP address, I need to change the database host here too.

From here onwards, I am using the MySQL query browser for editing the WordPress database options table. As mentioned before, you can also do this through phpmyadmin or through the MySQL command prompt or any other tool which lets you connect either locally or remotely to your MySQL server. The MySQL query browser has been now replaced by the MYSQL Workbench, but is still easier to use that the more comprehensive MySQL Workbench.

Connect to the server with a user account that has read/write access to the database.

Click to Enlarge

In the right pane under your database name, expand wp_options and then option_value
Click on the Execute button
Click on the Edit button
Change the option_value for option_name siteurl to whatever your new site URL is (again in my case, it is an IP address. You will need to change this only if you are changing your domain name too. If you are shifting hosting, you will need to change only the database server name as shown in the first screenshot)

Now you need to scroll down until you come to the option_name home.

Click to Enlarge

For the option_name home, repeat what you did in Step 4 above, and then click on the Apply Changes button (next to the Edit button) to save your changes.

This should get your WordPress blog up and running again

Solved: Error accessing a shared folder between two Windows computers

Benjy — Mon, 26 Jul 2010 06:20:19 +0000

This has been one of the oldest computer problems to plague me on a regular basis. On a peer to peer network, file and print sharing mostly works out of the box and computers can see and access each others shared resources without any problems. Sometimes something goes wrong and do what you might, the computers can no longer see each other when you search for them by their NETBIOS names. This can happen between any two windows versions. I have had endless nightmares when this happened between two Windows XP computers or between a Windows XP computer and another with Windows Vista. These are my conclusions after trying to tackle these problems over the last 5 to 6 years.

This can happen due to a few things that you can readily check, or you can have all of those things done and still stay stuck. It is a bit of a mystery, so I have usually solved it by workarounds, since many a time some computer just stops showing after an update, or a computer stops seeing the other computers. Searching Google shows that there are numerous other users suffering from the same problem. Here are some of the possible solutions.

One of the earliest solutions that worked for me was to have an account with the same username and password on both/all of the machines. This however worked well during the reign of Windows XP until service pack 1. Since Service Pack 2, this solution has worked only intermittently.
Some Internet security software written by paranoid people, block browsing your local network by default and you need to add an exception so that it doesn’t block your needed network resource.
Vista and Windows 7 also have this feature, and you can easily change this setting from their ‘Network and sharing center’
Your Computer browser service might be turned off or have been set up not to start automatically. You can check/change this from Control Panel > Administrative Tools > Services > Computer Browser
Enable NETBIOS over TCP/IP in your Network Adapter Properties. This can be done by viewing your Network adapter properties > TCP/IP (ver. 4) Properties > Advanced > WINS and selecting the radio button, “Enable NETBIOS over TCP/IP“
Plan B : Assign a static IP address to the computer hosting the share and access it by IP address rather than by its NETBIOS name.

If you get an authentication problem, then you can try the first solution. If that does not work, you can change settings for that particular share. (This cannot be done on the Home editions of Windows) This can be done from Control Panel > Administrative Tools > Local Security Policy .

In the Local Security Settings, under Security settings in the Left Pane… Local Policies > Security Options… double click ‘Network Access: Shares that can be accessed anonymously‘ . Here you can add your share name and click OK.

Enabling Anonymous access to Shares

Add your share name here

Solved: wordpress Fatal error: Allowed memory size of 33554432 bytes exhausted

Benjy — Tue, 06 Jul 2010 10:37:18 +0000

I’m trying out WordPress 3 as a platform for our office Intranet/Extranet. Setting it up on a LAMP installation of the 64 bit version of Ubuntu Server (Lucid), I could login only once to the wp-admin page. Whenever I tried again, I got the following error message:

wordpress Fatal error: Allowed memory size of 33554432 bytes exhausted (In xyz file)

Initially I thought that our server being low on memory was at fault since there were other applications running using the same MYSQL server. I unsuccessfully restarted the machine twice after shutting down some services like Zope hoping to free some memory and solve the problem.

The solution (after a bit of trial and error) is to ftp to your webserver, change over to your WordPress installation directory and edit the wp-config.php file.

Add the following line (on a new line) after the other lines beginning with “define”

define(‘WP_MEMORY_LIMIT’, ’64M’);

Hope this helped

Repairing Windows XP

Benjy — Sat, 10 Apr 2010 03:44:57 +0000

Ref : EHA/End User Self Help/Common IT Services setup

Windows XP is a bit low on security compared to Windows 7 and Vista -since both of them have implemented the sudo concept of Ubuntu and the Mac. Windows XP however did have the concept of standard/restricted users and users with full administrative rights but these features and the ‘run as’ command were seldom used as users continued to use the system with administrator rights like it was in the days of Windows ME, 98 and before.

The result of such use (or misuse) has been a highly unstable Operating system due to the fact that the system was now open to attack as every piece of malicious code would now run with administrative privileges and rule over all areas of the system. Add to this the common use of unlicensed software and the accompanying aversion to enabling automatic updates and unlicensed/un-updated anti virus software has led to un-patched security vulnerabilities and thus a ‘crash – format – reinstall – crash’ cycle keeps repeating. We are covering best practices for installing/maintaining Windows somewhere else, so we will have a look at how to get an existing system up and running with minimum effort. The last option is of course a format of the system -which is at best tedious as you will need to install drivers, setup printers and all application software which if done properly can take more than a day!

We will be looking at 5 different methods of recovery failing which you will be left with no option but to do a clean install of Windows and the tedious installation of drivers and application software

Running a disk check
System restore
Last Known good Configuration
Using the recovery console
Installation in repair mode

Running a disk check:

This is one of the easier things to do when you have intermittent problems on a Windows XP computer -You can perform a disk check by right clicking your C: drive in MY COMPUTER > Properties > Tools and clicking on the Check now button.

Personally though I prefer that the disk check takes place with the disks unmounted, so click on Start > Run and type in chkdsk c: /f and press Enter or click OK. It should prompt you to schedule the disk check to run when your computer restarts. Type y and press Enter and then restart your computer.

Using System restore:

There are two instances where you might want to run system restore:

When your computer is acting abnormally – Click on START > Programs > Accessories > System Tools > System Restore
When your computer fails to boot normally and you are forced to enter Safe mode, or you voluntarily choose to enter safe mode by pressing F8 during system startup.

In both the cases, you are presented with an option either to restore the system to a previous date (which are highlighted in bold) or to undo a previous restoration. However don’t expect to be able to restore your system to the date when you bought your Desktop. many times when critical changes have been made to your system, System restore might be unable to complete the restoration.

If you have a branded computer, it would have its own restore program in addition to Windows System restore -but be careful and back up your data! If you make a mistake and your computer gets restored to factory settings you might loose all your data!!

Last Known good configuration:

Sometimes you will be diverted to a black screen with lines of selectable text with options like Safe mode, Safe mode with networking etc… This usually happens if your computer failed to bootup the last time it was switched on.

This can also happen if you added some software or hardware which needed a change in some system settings. the solution here is pretty straight forward. Use the UP/Down arrow key to select Last Known Good Configuration (your most recent settings that worked) and press Enter.

Using the recovery console:

If you need to use this, then your computer never even reached the above black screen. Usually Windows will complain that some required files are missing and prompt you to use the windows CD.

Press any Key... If you don't, the installer on CD will transfer control over to the boot loader an the disk and boot normally

Press 'r' to enter the recovery console

You will be asked for the Administrator password. if you don’t know what it is, probably it was never set (a security risk). Press Enter and it should drop you to a command prompt. If it says the password is wrong and you don’t know what it is then you will have no option to do a repair install (which resets the administrator password or do a fresh install after a format!

Among others, there are two easy operations which a user can perform. I will not cover the other complicated commands.

Run a disk check/fix on computers that wont boot and…
Restore a damaged MBR (Master Boot Record)

I suggest you do both as there is no harm in doing so as it takes quite along time to get to the recovery console from CD, so if you tried the disk check and it didn’t work, you will need to re-do the boot from CD exercise and loose another 15-20 minutes.

The syntax for running the disk check is different in the recovery console. Chkdsk does not support the /f switch here. So to run a disk check from here, type the following at the command prompt…

chkdsk c: /p

You should see several lines scroll past saying that chkdsk id performing addition checking or recovery. Once chkdsk exits, write a new MBR to your primary hard disk using the command fixmbr

Restart the computer. If the problem was not resolved proceed to method No. 5

Installing Windows XP in repair mode:

A few important things to remember before carrying out this proceedure…

Back up all your data to a portable hard disk. if your system is unbootable, you may need to use a Live CD of Ubuntu/Knoppix to boot from and salvage your data -copy your whole user folder(s) under your Document and settings folder (Normally c:\documents and settings\username) Caution: If you ever renamed your username, your folder wil still be called by the old username -so be careful!!!
Keep your Windows product code ready You will be prompted for it during the install.
You will likely not be allowed to login (a Microsoft feature ) without activating your computer unless you have an Internet connection that is always on your network/modem supports DHCP! Keep the Microsoft activation phone numbers handy…get it from the net as the ones shown on screen might not work!!!

Select your Windows installation and press 'r' to repair

Follow the steps above till you come to the screen where you press ‘r‘ to enter the recovery console. Instead of pressing ‘r‘ here, press Enter to move to the next screen.

Accept the license agreement by pressing F8.

If windows was installed more than once (In different folders), then there will be as many Windows installations displayed here. Normally there would be only one installation as shown here and it will be selected by default. If it not, select the installation in question using the up/down arrow keys and press ‘r’ to begin the repair.

This will basically do a complete installation of Windows without disturbing your files, applications and settings.

After this is successfully done and Windows starts up normally, you might need to reinstall and update your anti virus software. All service packs and security updates for Windows also will need to be installed -even if you had done so before. Basically what we did was replace the Windows files in your Windows folder with the original (older) copies that came on your Windows CD.

If this fails, then you need to format and do a fresh install…. after copying your data of course

Solved: Cannot send outgoing Gmail in Evolution

Benjy — Sun, 07 Feb 2010 07:09:17 +0000

Last week I sent an email from my Gmail account which I had setup in Evolution on my home computer. Two days later I was surprised to find it still in the Outbox. I deleted the email as it was of no use now and made a mental To do to fix the problem as I was busy with other important things to be done -besides it still did work from my notebook -or I thought it did. A few days later I replied to a personal email from my notebook via my Gmail account and it is still lying in my outbox. This confirmed the fact that some settings had changed an Gmail’s end.

I was surprised at not receiving an email from Google to this effect, but gave them the benefit of doubt as I receive vast amounts of email and I could have easily have missed something from Gmail thinking it was some uninteresting email message like the various automated mails I receive from numerous bots whose websites I have subscribed to.Today I forwarded an interesting link to a friend and found it stuck in my outbox again which finally got my goat. A quick Google search revealed the usual mix of fixes -from the real ,to the imagined and to outright stupid. The secret to searching Google is the ability to distinguish between useful information and nonsense. As you get better at it, there will be a lot of links that you will not even bother to click.

Then there are those half truths…Someone said that Gmail had changed its outgoing port number and had not updated their How To docs…. believable, but its unlike Google not updating their docs and so I went to check out Gmail’s own help pages. I could not find Evolution listed there, but found some unintelligible hyper technical trash so I clicked on the link to Thunderbird and scrolled through the screenshots. The same stuff applies for Evolution as well.

Open Evolution and click on Edit > Preferences. Click on Mail Accounts and then select your Gmail account.

Click Edit and then on the sending email Tab. If you had a working Gmail account setup, You should see the following settings.

Server smtp.gmail.com:465

The checkbox on My Server requires authentication ticked

Security : Use secure connection SSL encryption

and your full email (gmail) address as your username

All you now need to do is change the port number for the outgoing smtp server smtp.gmail.com to 587 instead of 465 and then select TLS encryption instead of SSL! Now that I have done my good deed for the day, I can go ahead with my schedule for today afternoon/evening – A photo shoot of the Surajkund Annual Crafts mela (fair) The focus is on Rajasthan this year! Read more about it here

SMTP Settings for Gmail in Evolution

Tomcat6 error : jar not loaded. See Servlet Spec 2.3, section 9.7.2. Offending class: javax/servlet/Servlet.class

Benjy — Sun, 17 Jan 2010 10:01:46 +0000

This is with reference to my previous post here where I had to use a workaround to deploy a .war file for our Hospital Management Software. However as our date to go live approached, I was forced to do additional reading and experiment on the same and here are my conclusions. For your information, we are using Ubuntu Server 9.1 (Selecting LAMP + Tomcat Java Server from tasksel after the base install). the problem is still not sorted out but hopefully will soon be. The war file was deployed in /var/lib/tomcat6/webapps/. The logs are found at /var/log/tomcat6/

These are the error messages from the Catalinaxxxx.log.

jar not loaded. See Servlet Spec 2.3, section 9.7.2. Offending class: javax/servlet/Servlet.class

jar not loaded. See Servlet Spec 2.3, section 9.7.2. Offending class: javax/servlet/Servlet-api.class

The war file when copied to the /var/lib/tomcat6/ folder did not deploy. Trying to access it from the browser resulted in a 404 resource not found error

On Googling this error online, I came across the following description for Servlet Spec 2.3, section 9.7.2.

Servlet spec 2.3 sec 9.7.2 recommends … The classloader that a container uses to load a servlet in a WAR must allow the developer to load any resources contained in library JARs within the WAR following normal J2SE semantics using getResource. It must not allow the WAR to override J2SE or Java servlet API classes. It is further recommended that the loader not allow servlets in the WAR access to the web container’s implementation classes. It is recommended also that the application class loader be implemented so that classes and resources packaged within the WAR are loaded in preference to classes and resources residing in container-wide library JARs.

There were several solutions posted around the net and I am listing them out below.

The most common solution peddled online was to delete the offending file(s) from the WEB-INF/lib/ folder and ignore the error message. This seems to have worked for at least a few people. In our case, deleting the files did not enable the app to deploy properly. In fact we were rewarded with the -SEVERE: Error listenerStart error.
Other possible problems…That our developers had included the servlet-api.jar in the runtime library (in the .WAR) – and that it was only needed during build time. (They had)
Remove the jar from WEB-INF/lib, and update your build script to point to the new location. This too seems to be a possible solution, but this was out of my hands, and the solution was therefore passed on to the developers at JKT. (who just ignored it)
The problem could be that we had some other jar files which were not supposed to be in our war as they were bundled with the container(tomcat). To quote from another source, “It is a common mistake to keep these jars in WEB-INF/lib folder for compilation and ship it with the war file as these are only required for compiling your app (notably servlet-api.jar or j2ee.jar), but they should not be deployed as part of the webapp.” Another possibility -and conveyed to our developers with the same result as above… zzzz
“This is when you don’t have servlet-api.jar in class path. check that servlet-api.jar is in tomcat6/lib” -once again could be part of a possible problem but I have no access to the source code to verify this.
From yet another source, “the Servlet Spec 2.3, section 9.7.2 being referred to says that the Servlet Container (E.g. Tomcat) will supply the implementation class of the J2EE spec. The j2ee.jar (servlet-api.jar in our case?) in your WEB-INF/lib directory is trying to supply the same info. Having application specific implementations is a stability and security problem that is disallowed by the spec 2.3 and by Tomcat.” (Seems highly likely!!)
Here is an incomplete list of jar files (listed below) which are not supposed to be included in the war file but in the container’s lib directory.

j2ee.jar, jasper-*.jar, jsp-api.jar, rt.jar,tools.jar, servlet.jar, servlet-api.jar, xerce.jar, xerces.jar and xercesImpl.jar, jboss.jar, gwt-user.jar, gwt-dev-linux.jar, gwt-dev-windows.jar, standard.jar,

Unfortunately for us, I found information online which said that using servlet.jar & the servlet-api.jar for compiling is passe and compiling should preferably be done by using j2ee.jar instead. I’m still waiting for a reply from them (our developers) and have refused to deploy our software from a folder using a standalone copy of Tomcat. I received an email saying that there was no difference in both the methods -however they have been unable to deploy with the bundled version of Tomcat 6 in Ubuntu server, -how then were they able to compare both methods when they haven’t been able to set it up is anybody’s guess -divine guidance perhaps? Visions?

As a systems guy, for me the logic is simple. When anything is installed from a repository, the system administrator is assured of a centralized update and also automatically installed security updates (if opted for) -this is all the more important for us as we would ultimately be deploying around 20 servers running Ubuntu Server edition in primarily rural locations with no proper IT support -let alone Linux administrators. In addition to that we would use this system at least for the next 8-10 years so the technology used must be current. If I use the method propounded by our developers, (because doing it the optimal way would be additional work for them -incidentally they also prefer to always run as root) then I’d be running tomcat6 for the next 8-10 years and also the existing servlet-api.jar (among others). On the other hand, if I choose to run the Tomcat sever installed using tasksel or apt and the developers used the servlets bundled with it, then (its current servlet-api.jar in /usr/share/tomcat6/lib/ is a symlink to servlet-api-2.5.jar) I can get updates without breaking the app as any references to /usr/share/tomcat6/lib/servlet-api.jar gets pointed to its latest bug fixed/enhanced avatar, currently /usr/share/tomcat6/lib/servlet-api-2.5.jar. Since I am not a developer I and do not have access to the programming/compiling details or the source code, I cannot come to any definite conclusion for now but will append to this post as soon as I have more information on the same.

Posted some months later…

I tried a few experiments but couldn’t do much as I did not have access to the source code. The solution seems to build the package on the Eclipse Java EE IDE for Linux. It is available from the Eclipse Website. Download, extract and run.

Here is a screenshot of the screen after I imported the WAR file. It clearly states that ” Unselected libraries will be imported into the WEB-INF/lib folder as jar files” and that seems to be the most likely solution to the problem… build on Linux and select the libraries so that they don’t get included in your WEB-INF/lib folder! ( i.e. they won’t get included in your build and the copies that install along with Tomcat6 on the production server will be used) I’ve marked the libraries that are causing us issues with big red dots. The best way would be to select all so that Tomcat’s (on the production server) own libraries are used.

Importing a WAR into the Eclipse Java EE IDE on Linux

Posted a few months later:

In our case, what did work out finally was to set TOMCAT6_SECURITY=no in /etc/init.d/tomcat6 (The default is yes). Note: In Ubuntu Server 10.04 Lucid Lynx, the default is already set to no, so our HIS software works without a hitch. However the same error still shows up in the log files. It does not seem to affect any of its functionality though.

This is a stop gap “solution” from our software developers, JK Technosoft. who have designed the software. (knowing them, read that as a permanent solution) Can’t expect local Windows programmers to understand or care about the nitty-gritties of Linux security or trouble themselves even with Windows security for that matter. The last time I met them, their actions convinced me that they were not keen to explore the dilemma of their custom libraries residing in /WEB-INF/lib . I’m convinced that only third party libraries must make it into the /WEB-INF/lib folder -I could be wrong though, but if I was one of them, I’d avoid bundling the web libraries that are in any case automatically installed by default along with Tomcat. Its like bundling DLLs in a Windows application that Windows already has by default. The logic is simple, even in Tomcat on Linux as in Windows. The DLLs in the System folders in Windows are protected, secure and automatically updated, the dupes that the software developer has provided with the software that’s installing, might at best be outdated or at worst compromised! The only problem I see on using the system DLL’s would be that your app might break as part of a Windows or Linux update! Well that’s not supposed to happen (or happen only rarely) if you coded right in the first place!

In our case, the problem was point No. 6 among the possibilities listed above. However our developers continued to ignore our feedback. You can now read how the whole software story ended in this post. It wasn’t just a Tomcat problem, its just that they continued with their same callous attitude in all areas of their software.

remove the jar from WEB-INF/lib, and update your build script to point to the new location

Simplysimple.info » Troubleshooting

Recovering from a broken array : Ubuntu Software RAID

Related Posts

Taking screenshots with Lightscreen

Related Posts

Vista Graphic drivers for the Intel D845EPI/D845 GVSR Motherboard

Related Posts

[Solved] JRE persistent printing prompt

For Windows:

For Linux:

Related Posts

Moving your WordPress blog

Related Posts

Solved: Error accessing a shared folder between two Windows computers

Related Posts

Solved: wordpress Fatal error: Allowed memory size of 33554432 bytes exhausted

wordpress Fatal error: Allowed memory size of 33554432 bytes exhausted (In xyz file)

define(‘WP_MEMORY_LIMIT’, ’64M’);

Related Posts

Repairing Windows XP

Running a disk check:

Using System restore:

Last Known good configuration:

Using the recovery console:

Installing Windows XP in repair mode:

Related Posts

Solved: Cannot send outgoing Gmail in Evolution

Related Posts

Tomcat6 error : jar not loaded. See Servlet Spec 2.3, section 9.7.2. Offending class: javax/servlet/Servlet.class

Related Posts