cancel
Showing results for 
Search instead for 
Did you mean: 

ext3 block size change

SOLVED
Go to solution
Fred Ruffet
Honored Contributor

ext3 block size change

Hi all,

I have some servers initially created with a 512MB partition for the / filesystem. I am now asked to increase this size up to 35GB. Using LVM it's no big deal. lvextend and resize2fs. My concern is using resize2fs, I'm told this :
"Performing an on-line resize of /dev/vg00/lvol1 to 37093376 (1k) blocks."
I don't like the idea of the 1kB block for such a filesystem. Is there a way to increase this block size, without recreating the FS ?

Regards,

Fred
--

"Reality is just a point of view." (P. K. D.)
4 REPLIES
Matti_Kurkela
Honored Contributor
Solution

Re: ext3 block size change

Changing the block size would require completely regenerating all the filesystem metadata, and possibly moving files to conform to new block boundaries.

As far as I know, there is no way to change the block size without backing up all data, running "mke2fs -b ..." and restoring.

But you said this is about extending the root filesystem. I guess you're maintaining a separate /usr and /var to allow your / to be that small...

As you no doubt know, it would be a good idea to keep applications on their own filesystem(s): it makes upgrading the OS to a newer major release a lot easier.

Linux is less sensitive to the root filesystem filling up to 100% than HP-UX, but allowing random applications and data to be installed on the root filesystem is still a bad idea.

So instead of expanding your root filesystem, you should be trying to identify the paths needing extra capacity, and offering to create and mount extra filesystem(s) there.

In short, the request to extend the root filesystem from 512 MB to 35 GB sounds like whoever is supposed to plan the application installation is either not doing the job right is unaware of the best practices in a Unix-type system. Some gentle education might be needed...

MK
MK
Fred Ruffet
Honored Contributor

Re: ext3 block size change

Matti,

I'm aware of this considerations. But I have lost this battle.

For you to understand the whole thing : I was asked some days ago to install some servers on Virtual machines. My client planned to use them as development I installed those servers the good way : 512MB for /, 1GB for /var, 512MB for /tmp, 100MB for /boot, 3GB for /usr and what's left for user needs. All on LVM except /boot and so on. No X server running, no unnecessary package. Their application where supposed to be developed in /opt and they have sources and such things in /home. So I started with some space on those FS, planning to extend them at will.

Then, development manager explained to his boss (which is my client) that his applications may have to be installed on /usr, that they may log in /var, that they may need extra space on /tmp... and so on.

We had several explanations but all I managed to save was separation of /var and /tmp, and the insurance that when we will come to test and production servers, we will separate all as I did in first place.

So for now, I am increasing /, and reboot in single mode in order to move /usr to this ugly-big-unique /. I don't like it. I don't have choice.

But your answer about my primary question answers it perfectly : I can't increase block size. I may consider this as a revenge when they will complain about performance issue ;)

Regards,

Fred
--

"Reality is just a point of view." (P. K. D.)
TwoProc
Honored Contributor

Re: ext3 block size change

Fred, my problem with this thing of being setup with separate filesystems in production is that it will never happen. I mean really, are they going to *help you size the file systems correctly after go-live* ? And, they are OK with the unplanned downtime as a result of not knowing/understanding the sizing issues? Your boss gonna be OK with that? The end users gonna be OK with that?
We are the people our parents warned us about --Jimmy Buffett
Fred Ruffet
Honored Contributor

Re: ext3 block size change

John,

These are different servers. I mean, they will have servers dedicated for development (those I described), servers for single tests, servers for qualification, servers for beta-test and servers for production. So yes, they won't have the same FS layout for development and other parts. I can understand that developpers want to install different postgres versions and try devs rapidly without having to make the whole setup to have it strictly the same as in production. Developpers won't have their hands on other servers where needs and purposes will not be the same. We will have strictly controled servers on other platforms. And I won't let it go on those. trust me ;) I like to promote Linux as production servers and I will know how to argue and show them how and why other servers will have to be set up correctly.

Regards,

Fred
--

"Reality is just a point of view." (P. K. D.)