dohertj2/CBDDC
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Reformat/cleanup
All checks were successful
NuGet Package Publish / nuget (push) Successful in 1m10s
Details

Browse Source
This commit is contained in:
Joseph Doherty
2026-02-21 07:53:53 -05:00
parent c6f6d9329a
commit 7ebc2cb567
160 changed files with 7258 additions and 7262 deletions
Download Patch File Download Diff File Expand all files Collapse all files

56
src/ZB.MOM.WW.CBDDC.Hosting/HealthChecks/CBDDCHealthCheck.cs
View File

@@ -1,8 +1,3 @@
using System;
using System.Collections.Generic;
using System.Linq;
using System.Threading;
using System.Threading.Tasks;
using Microsoft.Extensions.Diagnostics.HealthChecks;
using ZB.MOM.WW.CBDDC.Core.Storage;
using ZB.MOM.WW.CBDDC.Hosting.Configuration;
@@ -10,17 +5,17 @@ using ZB.MOM.WW.CBDDC.Hosting.Configuration;
namespace ZB.MOM.WW.CBDDC.Hosting.HealthChecks;
/// <summary>
/// Health check for CBDDC persistence layer.
/// Verifies that the database connection is healthy.
/// Health check for CBDDC persistence layer.
/// Verifies that the database connection is healthy.
/// </summary>
public class CBDDCHealthCheck : IHealthCheck
{
private readonly IOplogStore _oplogStore;
private readonly IPeerOplogConfirmationStore _peerOplogConfirmationStore;
private readonly CBDDCHostingOptions _options;
private readonly IPeerOplogConfirmationStore _peerOplogConfirmationStore;
/// <summary>
/// Initializes a new instance of the <see cref="CBDDCHealthCheck"/> class.
/// Initializes a new instance of the <see cref="CBDDCHealthCheck" /> class.
/// </summary>
/// <param name="oplogStore">The oplog store used to verify persistence health.</param>
/// <param name="peerOplogConfirmationStore">The peer confirmation store used for confirmation lag health checks.</param>
@@ -31,16 +26,17 @@ public class CBDDCHealthCheck : IHealthCheck
CBDDCHostingOptions options)
{
_oplogStore = oplogStore ?? throw new ArgumentNullException(nameof(oplogStore));
_peerOplogConfirmationStore = peerOplogConfirmationStore ?? throw new ArgumentNullException(nameof(peerOplogConfirmationStore));
_peerOplogConfirmationStore = peerOplogConfirmationStore ??
throw new ArgumentNullException(nameof(peerOplogConfirmationStore));
_options = options ?? throw new ArgumentNullException(nameof(options));
}
/// <summary>
/// Performs a health check against the CBDDC persistence layer.
/// Performs a health check against the CBDDC persistence layer.
/// </summary>
/// <param name="context">The health check execution context.</param>
/// <param name="cancellationToken">A token used to cancel the health check.</param>
/// <returns>A <see cref="HealthCheckResult"/> describing the health status.</returns>
/// <returns>A <see cref="HealthCheckResult" /> describing the health status.</returns>
public async Task<HealthCheckResult> CheckHealthAsync(
HealthCheckContext context,
CancellationToken cancellationToken = default)
@@ -58,15 +54,18 @@ public class CBDDCHealthCheck : IHealthCheck
var peersWithNoConfirmation = new List<string>();
var laggingPeers = new List<string>();
var criticalLaggingPeers = new List<string>();
var lastSuccessfulConfirmationUpdateByPeer = new Dictionary<string, DateTimeOffset?>(StringComparer.Ordinal);
var lastSuccessfulConfirmationUpdateByPeer =
new Dictionary<string, DateTimeOffset?>(StringComparer.Ordinal);
var maxLagMs = 0L;
var lagThresholdMs = Math.Max(0, _options.Cluster.PeerConfirmationLagThresholdMs);
var criticalLagThresholdMs = Math.Max(lagThresholdMs, _options.Cluster.PeerConfirmationCriticalLagThresholdMs);
long lagThresholdMs = Math.Max(0, _options.Cluster.PeerConfirmationLagThresholdMs);
long criticalLagThresholdMs =
Math.Max(lagThresholdMs, _options.Cluster.PeerConfirmationCriticalLagThresholdMs);
foreach (var peerNodeId in trackedPeers)
foreach (string peerNodeId in trackedPeers)
{
var confirmations = (await _peerOplogConfirmationStore.GetConfirmationsForPeerAsync(peerNodeId, cancellationToken))
var confirmations =
(await _peerOplogConfirmationStore.GetConfirmationsForPeerAsync(peerNodeId, cancellationToken))
.Where(confirmation => confirmation.IsActive)
.ToList();
@@ -83,19 +82,14 @@ public class CBDDCHealthCheck : IHealthCheck
.ThenBy(confirmation => confirmation.ConfirmedLogic)
.First();
var lagMs = Math.Max(0, localHead.PhysicalTime - oldestConfirmation.ConfirmedWall);
long lagMs = Math.Max(0, localHead.PhysicalTime - oldestConfirmation.ConfirmedWall);
maxLagMs = Math.Max(maxLagMs, lagMs);
lastSuccessfulConfirmationUpdateByPeer[peerNodeId] = confirmations.Max(confirmation => confirmation.LastConfirmedUtc);
lastSuccessfulConfirmationUpdateByPeer[peerNodeId] =
confirmations.Max(confirmation => confirmation.LastConfirmedUtc);
if (lagMs > lagThresholdMs)
{
laggingPeers.Add(peerNodeId);
}
if (lagMs > lagThresholdMs) laggingPeers.Add(peerNodeId);
if (lagMs > criticalLagThresholdMs)
{
criticalLaggingPeers.Add(peerNodeId);
}
if (lagMs > criticalLagThresholdMs) criticalLaggingPeers.Add(peerNodeId);
}
var payload = new Dictionary<string, object>
@@ -108,18 +102,14 @@ public class CBDDCHealthCheck : IHealthCheck
};
if (criticalLaggingPeers.Count > 0)
{
return HealthCheckResult.Unhealthy(
$"CBDDC is unhealthy. Critical lag detected for {criticalLaggingPeers.Count} tracked peer(s).",
data: payload);
}
if (peersWithNoConfirmation.Count > 0 || laggingPeers.Count > 0)
{
return HealthCheckResult.Degraded(
$"CBDDC is degraded. Lagging peers: {laggingPeers.Count}, unconfirmed peers: {peersWithNoConfirmation.Count}.",
data: payload);
}
return HealthCheckResult.Healthy(
$"CBDDC is healthy. Latest timestamp: {localHead.PhysicalTime}.",
@@ -129,7 +119,7 @@ public class CBDDCHealthCheck : IHealthCheck
{
return HealthCheckResult.Unhealthy(
"CBDDC persistence layer is unavailable",
exception: ex);
ex);
}
}
}
}