very weird situations with SSD drives
I have run to a very weired-fustrated situation and i started a topic on CentOS however i think this is quite interesting wridle so i though to share it here and listen what you guys think.
I did solve the issues but funny thing is that the cause of the problem is still "undiscovered". How is that possible...keep reading. I am facing a very "nasty" situation where it is very tough to conclude / figure out where the problem might be coming from. I have setup two centos 5.9 systems which do clustering. The way i did it was with 30GB SSD drives where each disk of each system is: hda-- -------hda1=/boot -------hda2=/ -------hda3= cryptsetup partition -> drbd MOBOS: a) branchA b) MakerX the hda3 partition is unlocked on startup and then it is used as a drbd block device which is synchronized with the peer system. The problem is that i get various error messages on 2-3 different installs (either imaged or fresh): 1) I/O errors 2) DRQ errors 3) Some times data on the hda3 partition are scrambled (ascii chars appear like binary) 4) EXT3-fs error (device dm-0): ext3_lookup: unlinked inode 944717 in dir #944705 In the beginning i thought it could be a compatibility issue between the SSD and the MOBO. I changed the mobo (from branchB to makerX) and imaged the SSD disk from a backup image i had which was specifically for MakerX. All was working well until i made a reboot and then i noticed again that some files on the hda3 partition were not as they should. I decided to delete all files from the relevant partition and copy them back but when i deleted the files i then received the following error: Code:
EXT3-fs error (device dm-0): ext3_free_blocks: Freeing blocks not in datazone - block = 3681081083, count = 1 I am having hard time even to narrow down the possibilities which could cause the problem. I tried different motherboard with different SSD I tried same motherboard with three different different SSDs (mobos are embedded systems) but the errors are still hitting on random moments... On my last try i got again the below errors: Code:
EXT3-fs error (device dm-0): ext3_free_blocks: Freeing blocks not in datazone - block = 4040611822, c1 |
The issues disapeared since i started using LVMs . No more inodes errors or scrumbled data. However, could anyone have any logical explenation to this?
Why when creating a standar ext3 dos partion, encrypt it with LUKS and then use it as drbd block device you get such problems BUT if you just make an LVM ext3 partition and follow same procedure everything works fine? Could this be because LVM might use large "logical" block/clusters? I was talking with a friend and explainded what has sorted my problem and he asked me "What led you to try and use LVM? What logic did you think and made that move?" .. be honest i wasnt sure....it as just a speculation for me ...which worked |
very weird situations with SSD drives
have you tested the same images on non solid state drives?
|
Quote:
|
All times are GMT -5. The time now is 12:38 PM. |