[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: AW: Router Sharding

The discussion on routing has been improved in the latest 3.3 docs. Please take a look and see if it is helpful.

With sharding each route gets one or more labels and each router (shard) has a selector that selects a set of routes by label. So it is up to the user/admin to assign labels to the routes and then to select a set of routes on each router (shard). By default a router selects all routes, so when sharding is used all routers (including the default router) will need to have a selector.

ipf (VIP using VRRP) deployment selects a set of nodes on which to present the supplied set of ip addresses. At any point in time one of the nodes is the master and it receives the packets to the IP address. So a single IP address can be set in DNS with knowledge that it will be serviced by one of the nodes. This implies that the application has replicas running on all of the nodes in the set.


On 09/26/2016 04:53 PM, Aleksandar Lazic wrote:



I agree with you, and I have tried to contribute to the doc but that’s wasn’t an easy task so I stopped.

Maybe I was also to naïve so blame me that I have stopped contribution.


@1: Currently that’s not possible you will need to add for every route the label for the dedicate router.


‘oc create route …’


have no options to set labels you will need to use


oc expose service ... --labels='router=one' --hostname='...'


or you can use the labels in the webconsole.


Oh and by the way the default router MUST also have ROUTE_LABELS if you don’t want to expose all routes to the default router.


@2: you will need the new template from OCP 3.3 there are additional env variables necessary to be able to use more the none router on the same node.





and you need to add on the router nodes in the iptables chain ‘OS_FIREWALL_ALLOW’ the additional ports.


@3: This would be a little bit tricky on the same node due to the fact that the




only handle one config file. Maybe there is a way with *VIPS but I have never tried this.






Von: users-bounces lists openshift redhat com [mailto:users-bounces lists openshift redhat com] Im Auftrag von Srinivas Naga Kotaru (skotaru)
Gesendet: Montag, 26. September 2016 21:31
An: Andrew Lau <andrew andrewklau com>; users lists openshift redhat com
Betreff: Re: Router Sharding



Current sharding documentation is very high level, doesn’t cover step by step actual real world use cases.


Anyway, I was succeeded to create 2 shards. Lot of questions on this topic on how to proceed next …


1.      How to tell a project that all apps created on this project should use router #1 or router #2?

2.      Now we have 3 routers (default created as part of installation + additional 2 routers created). How the ports work? 80, 443 & 1936 assigned to default router. I changed ports to 81/444/1937 and 82/445/1938 to respectively shad #1 #2. These ports open automatically or explicit action required?

3.      Ipfailover (floating VIP) bound to default router. Do we need to create additional IP failover pods with different IP’s and match to shad #1 and #2? Or can we share same IP failover pods with single floating VIP to newly created shad’s as well?



Srinivas Kotaru


From: Andrew Lau <andrew andrewklau com>
Date: Friday, September 23, 2016 at 7:41 PM
To: Srinivas Naga Kotaru <skotaru cisco com>, "users lists openshift redhat com" <users lists openshift redhat com>
Subject: Re: Router Sharding


There are docs here:



On Sat, 24 Sep 2016 at 06:13 Srinivas Naga Kotaru (skotaru) <skotaru cisco com> wrote:

Just saw 3.3 features blog




We’re rethinking of our cluster design and want to consolidate 1 cluster per data center. Initially we were planning off 2 cluster per data center to server internal and external traffic dedicated to its own cluster.


Consolidating to a single cluster per DC will offer multiple advantages to us.  We currently running latest 3.2.1 release


Router Sharding is available in 3.2.x branch or need to wait for 3.3? I was thinking this feature has been available from 3.x onwards as per documentation available. Not sure what is mean for upcoming 3.3.


We really want to take advantage of this feature and test ASAP. Current documentation is not clear or explains only high level.


Can you help me or point to right documentation which explains step by steps to test this feature?


Can we control routes at project level so that clients wont modifies to move their routes from prod to non-prod or internal to external routers?



Srinivas Kotaru

users mailing list
users lists openshift redhat com

users mailing list
users lists openshift redhat com

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]