Re: [ceph-users] Problem making RadosGW dual stack

2017-10-31 Thread Wido den Hollander

> Op 30 oktober 2017 om 17:42 schreef alastair.dewhu...@stfc.ac.uk:
> 
> 
> Hello
> 
> We have a dual stack test machine running RadosGW.  It is currently 
> configured for IPv4 only.  This is done in the ceph.conf with:
> rgw_frontends="civetweb port=443s 
> ssl_certificate=/etc/grid-security/hostcertkey.pem num_threads=2000 
> access_log_file=/var/log/ceph/ceph-client.rgw.ceph-test-gw691-access.log 
> error_log_file=/var/log/ceph/ceph-client.rgw.ceph-test-gw691-error.log"
> 
> According to the civetweb documentation we can specify for it to listen to 
> both IPv4 and IPv6 addresses by doing: +443s or 443s,[::]:443s.  
> Unfortunately when we try either of these configurations (and a range of 
> other things we tried in desperation) the service fails to start and we see 
> this in the log:
> 2017-10-30 15:58:47.727956 7f7b2218fc80  0 ceph version 11.2.1 
> (e0354f9d3b1eea1d75a7dd487ba8098311be38a7), process radosgw, pid 12640
> 2017-10-30 15:58:47.994738 7f7b2218fc80  0 WARNING: skipping unknown 
> framework: 443s
> 2017-10-30 15:58:47.994789 7f7b2218fc80  0 starting handler: civetweb
> 2017-10-30 15:58:47.994875 7f7b2218fc80  0 civetweb: 0x7f7b22b2cb20: Cannot 
> add SSL socket (entry 1). Is -ssl_certificate option set?
> 2017-10-30 15:58:47.994889 7f7b2218fc80 -1 ERROR: failed run
> 
> We have tested that individually both the IPv4 and IPv6 works (the service 
> starts and transfers work), so we believe the problem is with how ceph parses 
> the port setting.  We did consider the possibility of running two radosgw 
> instances on the same machine, but thought it more sensible to ask if anyone 
> else has had the same problem and if they managed to fix it?  
> 

iirc just setting [::] would make the RGW bind to both v4 and v6.

I checked a RGW machine:

rgw frontends = civetweb port=[::]:7480

On Linux a v6 bind on all ports will also bind on v4.

Wido

> Thanks
> 
> Alastair
> ___
> ceph-users mailing list
> ceph-users@lists.ceph.com
> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
___
ceph-users mailing list
ceph-users@lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com


Re: [ceph-users] Problem making RadosGW dual stack

2017-10-30 Thread Robin H. Johnson
On Mon, Oct 30, 2017 at 04:42:00PM +, alastair.dewhu...@stfc.ac.uk wrote:
> Hello
..
> We have tested that individually both the IPv4 and IPv6 works (the
> service starts and transfers work), so we believe the problem is with
> how ceph parses the port setting.  We did consider the possibility of
> running two radosgw instances on the same machine, but thought it more
> sensible to ask if anyone else has had the same problem and if they
> managed to fix it?  
This looks like but 20942:
http://tracker.ceph.com/issues/20942

I worked around the port part by only binding to a single port and doing
the rest of the work in the load balancer. Still not possible to use
commas in some other args.

-- 
Robin Hugh Johnson
Gentoo Linux: Dev, Infra Lead, Foundation Asst. Treasurer
E-Mail   : robb...@gentoo.org
GnuPG FP : 11ACBA4F 4778E3F6 E4EDF38E B27B944E 34884E85
GnuPG FP : 7D0B3CEB E9B85B1F 825BCECF EE05E6F6 A48F6136
___
ceph-users mailing list
ceph-users@lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com


Re: [ceph-users] Problem making RadosGW dual stack

2017-10-30 Thread Denes Dolhay

Hello,


I'm not sure and not sure where, but I think I red that you can specify 
80+443s to accomplish this



Would you please try that?


Kind regards,

Denes.


On 10/30/2017 05:42 PM, alastair.dewhu...@stfc.ac.uk wrote:

Hello

We have a dual stack test machine running RadosGW.  It is currently configured 
for IPv4 only.  This is done in the ceph.conf with:
rgw_frontends="civetweb port=443s ssl_certificate=/etc/grid-security/hostcertkey.pem 
num_threads=2000 access_log_file=/var/log/ceph/ceph-client.rgw.ceph-test-gw691-access.log 
error_log_file=/var/log/ceph/ceph-client.rgw.ceph-test-gw691-error.log"

According to the civetweb documentation we can specify for it to listen to both 
IPv4 and IPv6 addresses by doing: +443s or 443s,[::]:443s.  Unfortunately when 
we try either of these configurations (and a range of other things we tried in 
desperation) the service fails to start and we see this in the log:
2017-10-30 15:58:47.727956 7f7b2218fc80  0 ceph version 11.2.1 
(e0354f9d3b1eea1d75a7dd487ba8098311be38a7), process radosgw, pid 12640
2017-10-30 15:58:47.994738 7f7b2218fc80  0 WARNING: skipping unknown framework: 
443s
2017-10-30 15:58:47.994789 7f7b2218fc80  0 starting handler: civetweb
2017-10-30 15:58:47.994875 7f7b2218fc80  0 civetweb: 0x7f7b22b2cb20: Cannot add 
SSL socket (entry 1). Is -ssl_certificate option set?
2017-10-30 15:58:47.994889 7f7b2218fc80 -1 ERROR: failed run

We have tested that individually both the IPv4 and IPv6 works (the service 
starts and transfers work), so we believe the problem is with how ceph parses 
the port setting.  We did consider the possibility of running two radosgw 
instances on the same machine, but thought it more sensible to ask if anyone 
else has had the same problem and if they managed to fix it?

Thanks

Alastair
___
ceph-users mailing list
ceph-users@lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com



___
ceph-users mailing list
ceph-users@lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com


[ceph-users] Problem making RadosGW dual stack

2017-10-30 Thread alastair.dewhurst
Hello

We have a dual stack test machine running RadosGW.  It is currently configured 
for IPv4 only.  This is done in the ceph.conf with:
rgw_frontends="civetweb port=443s 
ssl_certificate=/etc/grid-security/hostcertkey.pem num_threads=2000 
access_log_file=/var/log/ceph/ceph-client.rgw.ceph-test-gw691-access.log 
error_log_file=/var/log/ceph/ceph-client.rgw.ceph-test-gw691-error.log"

According to the civetweb documentation we can specify for it to listen to both 
IPv4 and IPv6 addresses by doing: +443s or 443s,[::]:443s.  Unfortunately when 
we try either of these configurations (and a range of other things we tried in 
desperation) the service fails to start and we see this in the log:
2017-10-30 15:58:47.727956 7f7b2218fc80  0 ceph version 11.2.1 
(e0354f9d3b1eea1d75a7dd487ba8098311be38a7), process radosgw, pid 12640
2017-10-30 15:58:47.994738 7f7b2218fc80  0 WARNING: skipping unknown framework: 
443s
2017-10-30 15:58:47.994789 7f7b2218fc80  0 starting handler: civetweb
2017-10-30 15:58:47.994875 7f7b2218fc80  0 civetweb: 0x7f7b22b2cb20: Cannot add 
SSL socket (entry 1). Is -ssl_certificate option set?
2017-10-30 15:58:47.994889 7f7b2218fc80 -1 ERROR: failed run

We have tested that individually both the IPv4 and IPv6 works (the service 
starts and transfers work), so we believe the problem is with how ceph parses 
the port setting.  We did consider the possibility of running two radosgw 
instances on the same machine, but thought it more sensible to ask if anyone 
else has had the same problem and if they managed to fix it?  

Thanks

Alastair
___
ceph-users mailing list
ceph-users@lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com