Server crash with ascii NUL characters in syslog ( ^@^@^@... )

Question

I have some dedicated server hosted by a OVH (French service provider). OS : Ubuntu 12.04 x64

A few months ago, one of my server crashed. The only weird thing was some "ASCII NUL" caracters in the syslog :

^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@ ^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@

With the help of my service provider, we checked :

• The RAM
• The CPU
• The DISKS

Everything was ok, so my service provider recommended to change the motherboard of the server and update the kernel (which we did). But since, this server crashed two more times, with the same caracters in the syslog.

Without any more explanation, we decided to change this server (this is planned in a few weeks).

BUT, problem is, this night, this happened to another server. Same crash, same caracters in the syslog, no explanation.

Does anybody have any clue of what we should check ? Is it a hardware or a software problem ?

Solution

I'll share more widely the great answer given by @n-st:

NUL bytes in the syslog are a common effect of a crash that prevented the system from cleanly syncing and unmounting the filesystem. They don't give an indication of what actually triggered the crash.

Indeed, I have often seen that behavior after a server crash: those characters are NULL (\0) characters that may represent a recovered block that was filled with zeros by some recover process.

As for the cause of the crash, that's a whole different question - you'd need to provide way more information for a diagnostic to even begin. I would recommend opening a different question about this if you still have the problem.

Comments

[BurninLeo]

Did you find a solution for this problem? I currently suffer the same issue...

[BurninLeo]

Actually, I did not find a solution (on a virtual server). After a while and some (regular) updates from the stable-releases, the problem vanished...

[n.st]

NUL bytes in the syslog are a common effect of a crash that prevented the system from cleanly syncing and unmounting the filesystem. They don't give an indication of what actually triggered the crash.

[BurninLeo]

I have a similar problem. Neither a long line nor the encoding explains the NUL characters at the end of the syslog (copied the file to an external disc and opened it with SciTE, UTF-8 encoding).

[Signal15]

It sounds like you may be opening the UTF-8 encoded file in an editor that does not understand UTF-8 very well. However it could be the CRLF problem (the dos2unix and unix2dos commands may be helpful)

[n.st]

NUL bytes in the syslog are a common effect of a crash that prevented the system from cleanly syncing and unmounting the filesystem. They don't give an indication of what actually triggered the crash.

[Signal15]

@n.st What a great answer! :) You should put that one as a "answer"