Skip to content

Commit

Permalink
Formatting
Browse files Browse the repository at this point in the history
  • Loading branch information
steved0x committed Jun 29, 2016
1 parent 7716c13 commit 186e1d4
Showing 1 changed file with 7 additions and 8 deletions.
15 changes: 7 additions & 8 deletions articles/redis-cache/cache-administration.md
Original file line number Diff line number Diff line change
Expand Up @@ -24,7 +24,7 @@ This topic describes how to perform administration tasks such as rebooting and s

## Administration settings

The Azure Redis Cache **Administration** settings allow you to perform the following administrative tasks for your premium cache. To access them, click **Settings** or **All settings** from the Redis Cache blade.
The Azure Redis Cache **Administration** settings allow you to perform the following administrative tasks for your premium cache. To access administration settings, click **Settings** or **All settings** from the Redis Cache blade.

![Administration](./media/cache-administration/redis-cache-administration.png)

Expand All @@ -45,11 +45,10 @@ To reboot one ore more nodes of your cache, select the desired nodes and click *

The impact on client applications varies depending on the node(s) that you reboot.

- **Master** - When the master node is rebooted, Azure Redis Cache fails over to the replica node and promotes it to master, with only a short interval in which connections may fail to the cache.
- **Slave** - When the slave node is rebooted, there is light to no impact to cache clients.
- **Master** - When the master node is rebooted, Azure Redis Cache fails over to the replica node and promotes it to master. During this failover there may be a short interval in which connections may fail to the cache.
- **Slave** - When the slave node is rebooted, there is typically no impact to cache clients.
- **Both master and slave** - When both cache nodes are rebooted, all data is lost in the cache and connections to the cache fail until the primary node comes back online.
- **Node(s) of a premium cache with clustering enabled** - When you reboot the node(s) of a premium cache with clustering enabled, the behavior is the same as when you reboot the node of a non-clustered cache, except that the impact is limited to the specific shard of the cache.

- **Node(s) of a premium cache with clustering enabled** - When you reboot the node(s) of a premium cache with clustering enabled, the behavior is the same as when you reboot node(s) of a non-clustered cache, except that the impact is limited to the specific shard of the cache.


>[AZURE.IMPORTANT] Reboot is only available for Premium tier caches.
Expand All @@ -60,7 +59,7 @@ The impact on client applications varies depending on the node(s) that you reboo
- [Can I reboot the cache to clear client connections?](#can-i-reboot-the-cache-to-clear-client-connections)
- [Will I lose data from my cache if I do a reboot?](#will-i-lose-data-from-my-cache-if-i-do-a-reboot)
- [Can I reboot my cache using PowerShell, CLI, or other management tools?](#can-i-reboot-my-cache-using-powershell-cli-or-other-management-tools)
- [What pricing tiers can use the Reboot functionality?](#what-pricing-tiers-can-use-the-reboot-functionality)
- [What pricing tiers can use the reboot functionality?](#what-pricing-tiers-can-use-the-reboot-functionality)


### Which node should I reboot to test my application?
Expand All @@ -69,7 +68,7 @@ To test the resiliency of your application against failure of the primary node o

### Can I reboot the cache to clear client connections?

Yes, if you reboot the cache all client connections are cleared. This can useful in the case where all client connections are used up, for example due to a logic error or a bug. Each pricing tier has different client connection limits for the various sizes, and once these limits are reached, no more client connections are accepted. Rebooting the cache provides a way to clear all client connections.
Yes, if you reboot the cache all client connections are cleared. This can useful in the case where all client connections are used up, for example due to a logic error or a bug in the client application. Each pricing tier has different [client connection limits](cache-configure.md#default-redis-server-configuration) for the various sizes, and once these limits are reached, no more client connections are accepted. Rebooting the cache provides a way to clear all client connections.

### Will I lose data from my cache if I do a reboot?

Expand All @@ -79,7 +78,7 @@ If you reboot both the **Master** and **Slave** nodes all data in the cache (or

At this time you can't, but this functionality is coming soon.

### What pricing tiers can use the Reboot functionality?
### What pricing tiers can use the reboot functionality?

Reboot is available only in the premium pricing tier.

Expand Down

0 comments on commit 186e1d4

Please sign in to comment.