Skip to main content

Monitoring and metrics

Watch-ZrStoreMetrics shows a live, continuously updating view of a store's throughput and activity, broken down by node. Use it to confirm a store is serving traffic, to watch load during a test, and to spot an imbalance or a bottleneck during an incident.

Activate a workspace first (see Connect to a cluster).

Watch live metrics​

Pass the store name with -StoreName. The view redraws in place until you stop it with Ctrl+C.

Watch-ZrStoreMetrics -StoreName <StoreName> [-RefreshSec <int>]

To monitor a store, refreshing once per second (the default):

Watch-ZrStoreMetrics -StoreName TestStore

To slow the refresh to every three seconds — useful over a slow link or to reduce console churn:

Watch-ZrStoreMetrics -StoreName TestStore -RefreshSec 3

-RefreshSec accepts 1–60 seconds. By default the view shows the Zaris metric category; pass -Category All to show every category, or -ShowSparkline to add a per-metric sparkline of recent values.

tip

Run Watch-ZrStoreMetrics in a dedicated terminal. It takes over the console and redraws on every refresh, so keep a separate session open for other admin commands.

Read the metrics view​

The view lists each metric as a row (grouped by category) and each node as a column. Every node column is split into a rate and a total sub-column. A healthy, active store shows rising totals and non-zero rates across its nodes.

Clustron Zaris | TestStore

Category         Metric                 10.0.0.11-n0            10.0.0.12-n0
                                            rate          total     rate          total
--------------------------------------------------------------------------------------
Zaris            get                    2.1K/s        154.0K   2.0K/s        149.2K
Zaris            put                    980/s          74.2K   1.0K/s         76.0K
Zaris            remove                 210/s          12.3K   198/s          11.9K
Zaris            store.count            --              2.3M   --              2.3M

Each node column shows a rate and a total:

  • rate is the average operations per second over the recent sampling window.
  • total is the cumulative count since the node started.
  • A -- in the rate position marks a gauge metric (such as store.count) that has no rate; only the total is shown.

Node columns are labeled by node id — for example, 10.0.0.11-n0 — so you can tell which node each column reports.

Interpret what you see​

The live view answers three operational questions.

  • Is the store serving traffic? Rising totals and non-zero rates mean the store is active. Flat totals across every node mean no client load, or an idle store.
  • Is load balanced? Similar rates across node columns are healthy. One node carrying far more or far less than the others points to an uneven key distribution or a client connected to a single seed node.
  • Is there a bottleneck? Rates that stall while clients are still sending work, or one node lagging the rest, are early signs of a saturated node or a network problem.

Troubleshoot the metrics view​

SymptomCauseFix
"No managers connected" errorNo workspace is activeActivate a workspace, or pass -Managers
No rows or all-zero totalsThe store is stopped, or no client is sending trafficConfirm the store is Running with Get-ZrStore; generate load with client operations
Totals rise but a node column stays flatThat node is down or unreachableCheck the node with Get-ZrStore; start it if stopped
View not updatingClient activity has stopped, or connectivity to a manager droppedVerify clients are active and the managers are reachable

Next steps​