Azure CosmosDB MongoDB API单文档Upsert触发超时异常求助

我们所有基于Azure CosmosDB MongoDB API（无服务器4.2版本）的服务每日都会出现间歇性异常：即便写入压力较低，随机的upsert操作会突然卡顿1分钟，并抛出MongoDB.Driver.MongoExecutionTimeoutException: Operation exceeded time limit.，部分日期出现频率更高。

核心操作与环境 #

• 写入独立集合，无事务、无复杂操作，仅通过_id读取文档后调用ReplaceOneAsync执行全文档upsert
• 使用C# + MongoDB.Driver 2.23.1版本，核心代码片段：

public async Task UpsertAsync(T item)
{
    await (await _mongoCollection.Value).ReplaceOneAsync(
        i => i.Id == item.Id,
        item,
        new ReplaceOptions { IsUpsert = true },
        CancellationToken.None);
}

public async Task<T?> TryGetByIdAsync(string id)
{
    var res = await (await _mongoCollection.Value).FindAsync(f => f.Id == id);
    return res.FirstOrDefault();
}

// 实体基类
public class MongoCollectionItem : CollectionItem
{
    [BsonId]
    [JsonPropertyName("_id")]
    public string Id { get; init; } = Guid.NewGuid().ToString();
}

• 已启用可重试写入（retryable writes）及服务端重试（server side retry）
• 排除限流可能性（无服务器实例）

完整异常栈 #

Error processing job for serial xxxxxx.x.x.xxx: MongoDB.Driver.MongoExecutionTimeoutException: Operation exceeded time limit.
   at MongoDB.Driver.Core.WireProtocol.CommandUsingCommandMessageWireProtocol`1.ProcessResponse(ConnectionId connectionId, CommandMessage responseMessage)
   at MongoDB.Driver.Core.WireProtocol.CommandUsingCommandMessageWireProtocol`1.ExecuteAsync(IConnection connection, CancellationToken cancellationToken)
   at MongoDB.Driver.Core.Servers.Server.ServerChannel.ExecuteProtocolAsync[TResult](IWireProtocol`1 protocol, ICoreSession session, CancellationToken cancellationToken)
   at MongoDB.Driver.Core.Operations.RetryableWriteOperationExecutor.ExecuteAsync[TResult](IRetryableWriteOperation`1 operation, RetryableWriteContext context, CancellationToken cancellationToken)
   at MongoDB.Driver.Core.Operations.BulkUnmixedWriteOperationBase`1.ExecuteBatchAsync(RetryableWriteContext context, Batch batch, CancellationToken cancellationToken)
   at MongoDB.Driver.Core.Operations.BulkUnmixedWriteOperationBase`1.ExecuteBatchesAsync(RetryableWriteContext context, CancellationToken cancellationToken)
   at MongoDB.Driver.Core.Operations.BulkMixedWriteOperation.ExecuteBatchAsync(RetryableWriteContext context, Batch batch, CancellationToken cancellationToken)
   at MongoDB.Driver.Core.Operations.BulkMixedWriteOperation.ExecuteAsync(IWriteBinding binding, CancellationToken cancellationToken)
   at MongoDB.Driver.OperationExecutor.ExecuteWriteOperationAsync[TResult](IWriteBinding binding, IWriteOperation`1 operation, CancellationToken cancellationToken)
   at MongoDB.Driver.MongoCollectionImpl`1.ExecuteWriteOperationAsync[TResult](IClientSessionHandle session, IWriteOperation`1 operation, CancellationToken cancellationToken)
   at MongoDB.Driver.MongoCollectionImpl`1.BulkWriteAsync(IClientSessionHandle session, IEnumerable`1 requests, BulkWriteOptions options, CancellationToken cancellationToken)
   at MongoDB.Driver.MongoCollectionImpl`1.UsingImplicitSessionAsync[TResult](Func`2 funcAsync, CancellationToken cancellationToken)
   at MongoDB.Driver.MongoCollectionBase`1.ReplaceOneAsync(FilterDefinition`1 filter, TDocument replacement, ReplaceOptions options, Func`3 bulkWriteAsync)
   at COMPANYNAME.Infrastructure.Persistence.Mongo.MongoCollection`1.UpsertAsync(T item)
...

1. 检查无服务器实例的冷启动/缩放延迟 #

无服务器CosmosDB会根据负载自动调整实例状态，休眠唤醒（冷启动）或扩容时可能出现请求延迟：

• 查看CosmosDB门户监控指标：关注Serverless Instance Count、Provisioned Throughput、Request Charge的波动，确认超时时段是否对应实例状态变化。

2. 优化客户端重试与超时配置 #

默认重试策略可能未覆盖所有场景，可调整客户端配置：

• 自定义重试规则：

var settings = MongoClientSettings.FromConnectionString(connectionString);
settings.RetrySettings = new RetrySettings
{
    MaxRetryCount = 5,
    MaxRetryWait = TimeSpan.FromSeconds(30),
    RetryTimeout = TimeSpan.FromSeconds(60)
};
settings.RetryWrites = true;
var client = new MongoClient(settings);

• 添加合理的取消令牌，避免请求无限制挂起：

using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(30));
await collection.ReplaceOneAsync(..., cancellationToken: cts.Token);

3. 验证索引与文档大小 #

• 确认_id字段默认索引存在（通过db.collection.getIndexes()验证）
• 检查超时时段文档是否突然变大，大文档写入可能触发内部操作延迟。

4. 排查后台维护操作 #

CosmosDB后台维护（如索引重建、数据迁移）可能影响性能：

• 在CosmosDB门户活动日志中查看超时时段是否有后台操作记录
• 查看Index Transformation指标，确认是否有索引相关任务在运行。

5. 调整客户端连接池配置 #

连接池耗尽或配置不合理可能导致请求排队：

• 调整MongoClientSettings中的ConnectionPoolSettings，优化MaxConnectionPoolSize、MinConnectionPoolSize参数
• 监控客户端连接状态，排查是否存在连接泄漏。

6. 升级MongoDB.Driver版本 #

当前2.23.1版本可能存在兼容性问题，建议升级到最新稳定版，查看官方release notes是否有修复CosmosDB相关超时问题。

内容的提问来源于stack exchange，提问作者der.Schtefan