List:General Discussion« Previous MessageNext Message »
From:Dan Nelson Date:June 15 2001 5:57pm
Subject:Re: mysql died with innodb
View as plain text  
In the last episode (Jun 15), Heikki Tuuri said:
> Sven,
> 
> line 637 in mysql/innobase/os/os0file.c asserts that
> 
> TRUE == os_file_flush(file)
> 
> If you look at about line 557, you see how InnoDB does
> the file flush: the default is to call fdatasync.
>
> To solve the problem we should look at the NFS manuals about what
> Unix functions are available in it. Maybe there is some special
> NFS_fsync?

When fsync() is called on an NFS filehandle, the system flushes any
dirty buffers on the client, and if over an NFSv3 mount, commits the
writes.

Sven: did you get any "nfs server timeout" messages in /var/log/messages?

Heikki: how hard would it be to send the errno back to the client if
filesystem calls fail?  On regular MYISAM tables, the user sees an
error message (or something gets dumped to the error log) and the
server doesn't assert().

-- 
	Dan Nelson
	dnelson@stripped
Thread
mysql died with innodbSven Huster15 Jun
  • Why not use indexes?Andrey Kotrekhov15 Jun
Re:mysql died with innodbHeikki Tuuri15 Jun
  • Re: mysql died with innodbDan Nelson15 Jun
    • RE: mysql died with innodbSven Huster15 Jun
Re: mysql died with innodbHeikki Tuuri15 Jun
  • Re: mysql died with innodbDan Nelson15 Jun
Re: mysql died with innodbHeikki Tuuri15 Jun