dm-devel.redhat.com archive mirror
 help / color / mirror / Atom feed
* path_grouping_policy  group_by_serial  prio question
@ 2016-03-08 12:03 Bernd Broermann
  2016-03-08 12:48 ` Christophe Varoqui
  0 siblings, 1 reply; 2+ messages in thread
From: Bernd Broermann @ 2016-03-08 12:03 UTC (permalink / raw)
  To: device-mapper development

Hello,


Does "group_by_serial" activate the "low latency"  path ?

There are two servers which have a separate path to a EMC VPLEX box in 
different Data centers ( DC1 , DC2 ) .

Because of latency the paths are grouped by the serial number of the 
VPLEX box.




        DC1                DC2

    +------------+    +------------+
    | Server 1   |    | Server 2   |
    +------------+    +------------+
       |   \              /  |
       |     \          /    |
       |       \      /      |
       |         \  /        |
       |          X          |
       |        /   \        |
       |      /       \      |
       |    /           \    |
       |  /              \   |
    +------------+    +------------+
    |disk VPLEX A|    |disk VPLEX B|
    +------------+    +------------+


As I expected the paths are grouped and the path to the "nearest" disk( 
VPLEX) is active.

Is this an coincidence or does it work as designed ?

I didn't find something enlightening in the dm documentation.

Thank you for your help.

Bernd



partitial output of  "multipathd show config"
....
device {
     vendor "EMC"
     product "Invista"
     product_blacklist "LUNZ"
     path_grouping_policy group_by_serial
     getuid_callout "/lib/udev/scsi_id --whitelisted --device=/dev/%n"
     path_selector "round-robin 0"
     path_checker tur
     features "0"
     hardware_handler "1 emc"
     prio emc
     rr_weight uniform
     no_path_retry fail
     rr_min_io 1000
     rr_min_io_rq 1
     detect_prio yes
}

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: path_grouping_policy group_by_serial prio question
  2016-03-08 12:03 path_grouping_policy group_by_serial prio question Bernd Broermann
@ 2016-03-08 12:48 ` Christophe Varoqui
  0 siblings, 0 replies; 2+ messages in thread
From: Christophe Varoqui @ 2016-03-08 12:48 UTC (permalink / raw)
  To: Bernd Broermann; +Cc: device-mapper development


[-- Attachment #1.1: Type: text/plain, Size: 2098 bytes --]

Hi,

the path_grouping_policy is only responsible for the path grouping.
The "emc" prioritizer here, attributes a prio to each path.
Priorities of paths in the same group are aggregated into a pathgroup
priority.
The pathgroup with the highest priority is activated.

Best regards,
Christophe Varoqui
www.opensvc.com


On Tue, Mar 8, 2016 at 1:03 PM, Bernd Broermann <bernd@broermann.com> wrote:

> Hello,
>
>
> Does "group_by_serial" activate the "low latency"  path ?
>
> There are two servers which have a separate path to a EMC VPLEX box in
> different Data centers ( DC1 , DC2 ) .
>
> Because of latency the paths are grouped by the serial number of the VPLEX
> box.
>
>
>
>
>        DC1                DC2
>
>    +------------+    +------------+
>    | Server 1   |    | Server 2   |
>    +------------+    +------------+
>       |   \              /  |
>       |     \          /    |
>       |       \      /      |
>       |         \  /        |
>       |          X          |
>       |        /   \        |
>       |      /       \      |
>       |    /           \    |
>       |  /              \   |
>    +------------+    +------------+
>    |disk VPLEX A|    |disk VPLEX B|
>    +------------+    +------------+
>
>
> As I expected the paths are grouped and the path to the "nearest" disk(
> VPLEX) is active.
>
> Is this an coincidence or does it work as designed ?
>
> I didn't find something enlightening in the dm documentation.
>
> Thank you for your help.
>
> Bernd
>
>
>
> partitial output of  "multipathd show config"
> ....
> device {
>     vendor "EMC"
>     product "Invista"
>     product_blacklist "LUNZ"
>     path_grouping_policy group_by_serial
>     getuid_callout "/lib/udev/scsi_id --whitelisted --device=/dev/%n"
>     path_selector "round-robin 0"
>     path_checker tur
>     features "0"
>     hardware_handler "1 emc"
>     prio emc
>     rr_weight uniform
>     no_path_retry fail
>     rr_min_io 1000
>     rr_min_io_rq 1
>     detect_prio yes
> }
>
> --
> dm-devel mailing list
> dm-devel@redhat.com
> https://www.redhat.com/mailman/listinfo/dm-devel
>

[-- Attachment #1.2: Type: text/html, Size: 3562 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]



^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2016-03-08 12:48 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2016-03-08 12:03 path_grouping_policy group_by_serial prio question Bernd Broermann
2016-03-08 12:48 ` Christophe Varoqui

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).