Issue 241 in roster-dns-management: edge case: failed network file transfer to nameserver

0 views
Skip to first unread message

roster-dns...@googlecode.com

unread,
Apr 12, 2012, 5:06:26 PM4/12/12
to roster-iss...@googlegroups.com
Status: New
Owner: lsar...@gmail.com
CC: ar...@purdue.edu, jjwiel...@gmail.com
Labels: Type-Defect Priority-High Security

New issue 241 by lsar...@gmail.com: edge case: failed network file transfer
to nameserver
http://code.google.com/p/roster-dns-management/issues/detail?id=241

What steps will reproduce the problem?

1. Suppose there is a network glitch when roster is pushing data to a
nameserver which causes the file(s) to be truncated or corrupted. Suppose
the corrupted files are still readable by BIND and are read into a
production nameserver.


What is the expected output? What do you see instead?

A nameserver successfully reading corrupted or truncated zone files could
cause partial to catastrophic DNS outage.

Is there a way to autodetect failed data transfer and/or possibly automate
recovery?

What version of the product are you using? On what operating system?

Roster: Any version.
OS: RHEL6
BIND: 9.9.0

Please provide any additional information below.


roster-dns...@googlecode.com

unread,
May 15, 2012, 4:20:57 PM5/15/12
to roster-iss...@googlegroups.com
Updates:
Status: Accepted
Owner: ---

Comment #1 on issue 241 by sharrell...@gmail.com: edge case: failed network
The gist of this is needing to make the placement of files an atomic
operation that will complete regardless of network status.

Currently what would happen is that the files would be placed incompletely
but the bind server would never be restarted. This is not catastrophic but
it is less than ideal.

roster-dns...@googlecode.com

unread,
May 16, 2012, 3:00:39 PM5/16/12
to roster-iss...@googlegroups.com
Updates:
Cc: lsar...@gmail.com

Comment #2 on issue 241 by lsar...@gmail.com: edge case: failed network
When you say named would never be restarted, do you mean named crashes and
cannot be restarted because of corrupt zone files?

roster-dns...@googlecode.com

unread,
May 16, 2012, 3:07:11 PM5/16/12
to roster-iss...@googlegroups.com

Comment #3 on issue 241 by sharrell...@gmail.com: edge case: failed network
No I mean there is a series of commands and if the command before
restarting named which is placing the files fails then the rndc reload
command will not happen so bind will not be reloading bad data.

roster-dns...@googlecode.com

unread,
May 30, 2012, 10:50:56 AM5/30/12
to roster-iss...@googlegroups.com
Updates:
Labels: Release-0.17

Comment #4 on issue 241 by sharrell...@gmail.com: edge case: failed network
(No comment was entered for this change.)

roster-dns...@googlecode.com

unread,
Jul 30, 2012, 3:39:45 PM7/30/12
to roster-iss...@googlegroups.com
Updates:
Labels: -Release-0.17

Comment #5 on issue 241 by sharrell...@gmail.com: edge case: failed network

roster-dns...@googlecode.com

unread,
Jul 30, 2012, 3:40:45 PM7/30/12
to roster-iss...@googlegroups.com
Updates:
Owner: Stephen....@gmail.com

Comment #6 on issue 241 by sharrell...@gmail.com: edge case: failed network

roster-dns...@googlecode.com

unread,
Aug 3, 2012, 11:29:13 AM8/3/12
to roster-iss...@googlegroups.com
Updates:
Labels: Project-FilePushingRedesign

Comment #7 on issue 241 by sharrell...@gmail.com: edge case: failed network

roster-dns...@googlecode.com

unread,
Aug 3, 2012, 11:30:13 AM8/3/12
to roster-iss...@googlegroups.com
Updates:
Labels: -Project-FilePushingRedesign Release-FilePushingRedesign

Comment #8 on issue 241 by sharrell...@gmail.com: edge case: failed network

roster-dns...@googlecode.com

unread,
Aug 7, 2012, 5:42:39 PM8/7/12
to roster-iss...@googlegroups.com
Updates:
Owner: ---
Labels: -Release-FilePushingRedesign Release-0.18
Blockedon: roster-dns-management:317

Comment #9 on issue 241 by sharrell...@gmail.com: edge case: failed network

roster-dns...@googlecode.com

unread,
Jul 1, 2013, 12:27:14 PM7/1/13
to roster-iss...@googlegroups.com

Comment #10 on issue 241 by J.S.P...@gmail.com: edge case: failed
Under Linux, directory swapping is not atomic. However, symlink swapping
is. So, to make this atomic, we must make Roster set up the DNS servers to
use a symlink to the zone files directory, and then after that all exports
can and will be atomic.

--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings

roster-dns...@googlecode.com

unread,
Jul 19, 2013, 4:57:54 PM7/19/13
to roster-iss...@googlegroups.com
Updates:
Status: Fixed
Blockedon: -roster-dns-management:317

Comment #11 on issue 241 by J.S.P...@gmail.com: edge case: failed
Fixed in r1152.
Reply all
Reply to author
Forward
0 new messages