Spent the last few nights neck-deep in logs and traces, which got me thinking about a different kind of search: enterprise search for our own documentation, runbooks, and past incident threads. We trialed Searchable last quarter, and while the relevance was decent, their alerting and SLA dashboards felt like an afterthought.
For those evaluating in 2026, who are the real contenders? I'm looking at platforms that can handle:
* **Scale:** Indexing millions of technical documents, code snippets, and Slack/Teams histories.
* **Freshness:** Near real-time updates. If a runbook changes post-incident, search results must reflect it in minutes, not days.
* **Observability Integration:** This is key for me. Can I expose metrics like query latency, top failed searches, or index health to Prometheus? I need to build a dashboard to prove it's a critical service, not a black box.
From our research, the shortlist we're stress-testing includes:
* **Elasticsearch (self-managed / Cloud):** Still the beast. The control is unmatched, but the operational overhead is a real on-call consideration. You'll be building your own Grafana dashboards for cluster health.
```yaml
# Example Prometheus exporter for Elasticsearch metrics you'd care about
- job_name: 'elasticsearch'
static_configs:
- targets: ['es-node:9114']
relabel_configs:
- source_labels: [__address__]
target_label: instance
```
* **Algolia:** Developer experience is top-tier. Their analytics API is great, but you need to pipe that data elsewhere for correlation with other system metrics. Pricing scales with records and operations, which can get interesting with high-volume internal logs.
* **Azure Cognitive Search:** If you're already in the Azure ecosystem, the integration is seamless. Their AI-powered features for semantic search are compelling for messy, natural-language queries like "that time the database latency spiked last Tuesday."
* **Vespa:** The performance profile is impressive for hybrid search (text + metrics). It's complex to operate, but it feels like it was built for the kind of large, varied datasets we have in observability.
Anyone else running a side-by-side comparison, especially focusing on operational visibility and total cost of ownership at scale? I'm less interested in marketing claims and more in what metrics you're exporting and how you're alerting on search degradation.
zzz
Sleep is for the weak