T
T
tictac172019-11-12 09:21:20
linux
tictac17, 2019-11-12 09:21:20

Why is a broken zip archive created in a shared folder?

Faced such strangeness. There is a 1C server and a backup server (RK server). The 1C server is virtual, located on the Proxmox hypervisor. The RK server is physical, it has a shared folder for 1C copies. The hypervisor also has a shared folder for "just in case" duplicate copies of 1C databases (well, just the disks are large, there's nothing to score anyway). Every day at 00-00, the 1C server makes a backup copy of the databases in a zip archive and puts it in these shared folders. The Zip archive is created twice, the order is the hypervisor, the RK server. Zip-archive is created immediately in the shared folders.
Now the problem. In the folder from the hypervisor, the archives are intact, in the folder of the RK server - broken! And there and there the same Debian, the same version of Samba, even the configs are comparable, default. For me it looks very strange, especially when you consider that a lot of things are copied to the RK server, the same images of virtual machines - there is nothing more broken! Only the bases with 1C, or rather the bases in zip'e. I checked, maybe some other routine tasks intersect - but no! Where to dig?
PS: a simple self-written script is used for backup, it stops samba and apache (bases through it), mounts shared folders, zips databases there, umounts folders and starts samba and apache.

Answer the question

In order to leave comments, you need to log in

1 answer(s)
T
tictac17, 2019-11-13
@tictac17

The answer to my question has been found. I found this in the hypervisor logs:
e1000e 0000:00:19.0 eth0: Detected Hardware Unit Hang
That is, the network card simply turned off for a couple of seconds due to the fact that it went into a "stupor". Probably an old bug with power saving or TSO offload.
It's funny that the network card corrupted the information, but the IP / TCP checksums were correct. The same couple of thousand bytes that passed through the insides of the hypervisor and then to the network card were spoiled. The network card of the RK server accepted them as correct (the amounts are correct). By the way, the server is branded, so as not to be reproached for stunted iron.
I hope I wrote this text here not in vain and it will help someone with a similar problem.

Didn't find what you were looking for?

Ask your question

Ask a Question

731 491 924 answers to any question