Re: Cfengin problems when copying large amounts of data over the ne t

[Mark . Burgess]

That is a surprise -- I have never seen this error message occur. 
Are you using encryption to transport all these files?

Anyway watch this space, because there is a significant rewrite
of this code, which I am just testing now.

Mark


On  6 Feb, Daniel Riek wrote:
> Hi,
> 
> perhaps I missed something for the config, but I have a problem distributing
> larg amounts of data over a lan with cfengine 2a16.
> 
> We are using cfengine for the management of a server farm. On job here is 
> the distribution of software updates in form of rpm-packets and tarballs.
> The software lays in several subdirectories according to the system-class
> it should be installed on (eg. webservers, infrastructure servers, ...).
> If new software is copied, we use a define to have cfengine run rpm -U or
> -F over the directory. - So far this works as we intend.
> 
> One special case are the infrastructure servers. They are two in a redundant
> ha setup (infra1 and infra2). All management is done from infra1 and I tried
> to use cfengine to copy the software repository to infra2 when something 
> changes. Thus in the first run of cfagent on infra2 a lot of files is going
> to be copied (a whole RedHat RPM-tree, ao.).
> Now after some time of running and copying I will get something like the 
> following output for the rest of the files:
> cf:infra2: 
> /opt/alcove-iso/install-RPMS-new/RPMS/rwall-server-0.17-10.i386.rpm wasn't at 
> destination (copying)
> cf:infra2: Copying from 
> infra.skk-eb:/opt/alcove-iso/install-RPMS-new/RPMS/rwall-server-0.17-10.i386.rpm
> cf:infra2: WARNING: new file 
> /opt/alcove-iso/install-RPMS-new/RPMS/rwall-server-0.17-10.i386.rpm.cfnew 
> seems to have been corrupted in transit, aborting!
> cf:infra2: /opt/alcove-iso/install-RPMS-new/RPMS/rxvt-2.7.6-4.i386.rpm wasn't 
> at destination (copying)
> cf:infra2: Copying from 
> infra.skk-eb:/opt/alcove-iso/install-RPMS-new/RPMS/rxvt-2.7.6-4.i386.rpm
> cf:infra2: WARNING: new file 
> /opt/alcove-iso/install-RPMS-new/RPMS/rxvt-2.7.6-4.i386.rpm.cfnew seems to 
> have been corrupted in transit, aborting!
> ...
> 
> Then the .cfnew are purged and sometimes cfagent segfaults, sometimes just 
> runs through.
> Is it possible, that the checksum cache is exhausted, or something like
> that? The file copy itself actually works correctly for the md5sum on the 
> .cfnew that 
> is purged lateron is the same as the of the original file on the policyhost.
> 
> Another things was, that a15 sometimes segfaultet when trying to copy 
> symlinkgs during a recursive directory copy. But I did not try that
> with a16 now.
> 
> Any ideas?
> 
> Regards,
> 
> Daniel
> 



~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Work: +47 22453272            Email:  Mark.Burgess@iu.hio.no
Fax : +47 22453205            WWW  :  http://www.iu.hio.no/~mark
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~