Bug 1188397 - RadosGW documentation should not use /var/run/ceph for socket path
Summary: RadosGW documentation should not use /var/run/ceph for socket path
Keywords:
Status: CLOSED CURRENTRELEASE
Alias: None
Product: Red Hat Ceph Storage
Classification: Red Hat Storage
Component: Documentation
Version: 1.2.2
Hardware: Unspecified
OS: Unspecified
unspecified
unspecified
Target Milestone: pre-dev-freeze
: 1.2.3
Assignee: Nilamdyuti
QA Contact: Ian Colle
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2015-02-02 18:24 UTC by Alexandre Marangone
Modified: 2015-10-05 22:55 UTC (History)
7 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2015-10-05 22:55:02 UTC
Embargoed:


Attachments (Terms of Use)


Links
System ID Private Priority Status Summary Last Updated
Ceph Project Bug Tracker 9001 0 None None None Never

Description Alexandre Marangone 2015-02-02 18:24:37 UTC
Description of problem:

With RHEL and CentOS, the RGW is running as apache making it impossible to create a socket in /var/run/ceph.

Version-Release number of selected component (if applicable):
All

How reproducible:
All the time

Steps to Reproduce:
1. Follow http://ceph.com/docs/master/radosgw/config/#add-a-gateway-configuration-to-ceph on any RPM supported distro

Suggestion: 
Update doc so that "rgw socket path" and "FastCgiExternalServer" use a new directory for the socket (e.g. /var/run/radosgw/).

Link to doc pages referencing /var/run/ceph: 
- http://ceph.com/docs/master/radosgw/config/#create-a-gateway-configuration
- http://ceph.com/docs/master/radosgw/config/#add-a-gateway-configuration-to-ceph
- http://ceph.com/docs/master/radosgw/federated-config/

Comment 1 John Poelstra 2015-02-12 18:56:11 UTC
Also need a bug filed upstream

Comment 3 John Wilkins 2015-03-04 03:33:50 UTC
Assigned to Nilam.

Comment 5 Ken Dreyer (Red Hat) 2015-03-09 20:24:24 UTC
Alexandre, you're correct that the best solution is to have RGW use an entirely separate directory that's chown'd to Apache. I've proposed that upstream in a wip- branch in Git, at http://tracker.ceph.com/issues/9001

In the mean time, for RHCeph 1.2.x, we should just document the workaround of running chown by hand. It's a pain, and it may break "rpm -V ceph", but I don't think we have time to test moving the entire FCGI socket location at this point. There are probably other things that rely on the FCGI socket being in /var/run/ceph, and I'd really rather get this settled in upstream instead of deviating downstream at the last minute.

Proposed doc change @ https://gitlab.cee.redhat.com/jowilkin/red-hat-ceph-storage-object-gateway/merge_requests/3

Comment 7 Ian Colle 2015-03-10 15:22:24 UTC
Yep - it's there.

Comment 8 Ken Dreyer (Red Hat) 2015-03-12 21:55:03 UTC
After discussing a related issue (https://github.com/ceph/ceph/pull/3916) with dalgaaf upstream...

On RHEL 6, /var/run is on "/" (ext4) by default, so the "chown apache" operation will persist across reboots.

On RHEL 7, /var/run is a symlink to /run, which is tmpfs. This means that "chown /var/run/ceph" will not persist across reboots.

I think the correct solution for RHEL 7 is to create a tmpfiles.d entry so that systemd knows to make /var/run/ceph (ie /run/ceph) owned by apache.


Note You need to log in before you can comment on or make changes to this bug.