Joined October 2023
- Tweets 83
- Following 9
- Followers 53
- Likes 38
39 Photos and videos
Jun 12
#ClickHouse mistake 3: treating PRIMARY KEY like Postgres.
It doesn't enforce uniqueness. It's a sparse index for skipping data ranges.
The real lever is ORDER BY: lowest → highest cardinality, matching your filters.
Learn more 👇
glassflow.dev/blog/clickhous…
#DataAnalytics
22
Jun 10
We went through hundreds of Stack Overflow threads, GitHub issues, and Reddit posts about #ClickHouse.
The same 5 mistakes keep coming up. Every. Single. Time.
Here are the first two 🧵
1
47
Jun 10
Mistake 2: Using MergeTree because it's the default engine.
ClickHouse has many MergeTree variants for a reason:
Building CDC or event tracking? → ReplacingMergeTree
Pre-computing metrics? → AggregatingMergeTree & Mat Views
Choose the engine wisely before your first insert.
1
27
Jun 10
Full breakdown with SQL examples for both ( 3 more mistakes incoming):
glassflow.dev/blog/clickhous…
Dropping one mistake every 2 days. Follow to catch the rest.
7
GlassFlow retweeted
Best tool out there for Kafka to dedup
May 12
Stop writing glue code to connect #OTel to #ClickHouse. 🛑
GlassFlow now ingests OTLP data natively: deduplicating, masking PII, and delivering traces/logs/metrics query-ready.
✅ No #Kafka
✅ No custom transforms
✅ Just clean data
See how: glassflow.dev/blog/clean-enr…
#OpenTelemetry
1
2
27
Jun 2
Before v3.1.0, when a downstream stage couldn't keep up, GlassFlow had no coordinated way to slow down. Events would pile up, NATS memory would fill, pipeline would fail. That's fixed now. 🧵
1
1
22
Jun 2
Also in v3.2.0: OTLP receiver concurrency caps, operator reconciles 4 pipelines in parallel, and full backpressure signals from every component to the control plane.
1
1
7
May 28
Most teams default to OTel → Kafka → ClickHouse for observability.
Kafka is powerful. But is it always necessary?
We broke down when it's worth the overhead and when it isn't.
👉 glassflow.dev/blog/opentelem…
#ClickHouse #OpenTelemetry #DataEngineering #Kafka #OTel
1
39
May 22
ReplacingMergeTree deduplicates eventually in #ClickHouse.
Your counts might be wrong until the merge runs.
Fix it upstream with GlassFlow → glassflow.dev
#DataEngineering #GlassFlow #OpenSource #DataPipelines
51
May 20
If you're sending #OTel data to ClickHouse without a processing layer in between, you're probably storing 15–30% duplicate logs and fighting nested JSON schema issues.
We built GlassFlow to fix this:dedup, enrichment, & schema mapping for #ClickHouse.
→ glassflow.dev/use-cases/real…
33
May 19
Still looking for a Founding Engineer 🚀🚀
Not a "senior dev" role. Not a backlog ticket role. A "here's the problem, let's build for it" role.
→ Low-latency #eventstreaming infra
→ Billions of events
→ GlassFlow stack: #Go, #Kafka, #NATS
Apply here: join.com/companies/getglassf…
1
145
May 13
Your Kafka → #ClickHouse pipeline shouldn't be a 2 AM nightmare.
ReplacingMergeTree doesn't deduplicate on insert. It merges "eventually", leading in messy count()s
GlassFlow fixes this by handling dedup, PII masking, and more upstream
➡️ Details: glassflow.dev
#Kafka
106
May 12
Stop writing glue code to connect #OTel to #ClickHouse. 🛑
GlassFlow now ingests OTLP data natively: deduplicating, masking PII, and delivering traces/logs/metrics query-ready.
✅ No #Kafka
✅ No custom transforms
✅ Just clean data
See how: glassflow.dev/blog/clean-enr…
#OpenTelemetry
2
123
May 7
ClickHouse for #observability is great, until you have to deal with:
• Duplicate spans from collector retries
• PII leaking into trace attributes
New guide: #OTel Collector → GlassFlow → #ClickHouse
Dedup, PII masking, tail sampling.
🔗 Link in replies
#OpenTelemetry
1
43
May 5
Working with #ClickHouse and running into the "Too Many Parts" errors? Let the server do the heavy lifting with Asynchronous Inserts.
Check our article on Asynchronous Inserts in ClickHouse here: glassflow.dev/blog/asynchron…
#DataOps #Analytics #DataEngineering #RealTimeData
46