The ClickHouse team sponsored an in-person ClickHouse meetup on 8 June 2022 at the Startup Village in Amsterdam. This was an exciting event, as it was the first in-person ClickHouse community gathering since February 2020. Altinity naturally attended – Mikhail Filimonov and Robert Hodges traveled to Amsterdam. Here is our trip report.
The Amsterdam database community includes many ClickHouse users, and the meeting was correspondingly well attended. About 100 people were there including 20 or more from ClickHouse Inc, who arranged the event and have a great office in downtown Amsterdam. It was great to see the core committers in person again, including Alexey Milovidov, Sasha Sapin, Nicolai Kochetov, and many others. Many Altinity customers and friends were there as well.
Here’s a summary of the technical talks, which were streamed to YouTube. Descriptions include links to positions in the YouTube stream.
- Lorenzo Mangani of Uptrace andQXIP described Uptrace and Cloki, which store telemetry data, traces, and services logs in ClickHouse. Among other innovations Cloki implements a gateway that translates PromQL/LogQL to ClickHouse SQL. You can look at the data using Grafana’s Loki data source, which is pretty cool.
- Next, there were three lightning talks.
- Ilya Yatsishin of ClickHouse Inc talked about ClickHouse benchmarks. Most significant point: AWS Graviton is fast. Altinity demonstrated that a while back, but Ilya’s numbers look even better. ClickHouse support for ARM is getting good.
- Ryadh Dahimen, likewise of ClickHouse Inc, presented a history of service logs and their search for a home. The moral of the story: ClickHouse is a great store for logs.
- Mikhail Filimonov of Altinity explained a common – and disastrous – anti-pattern in ClickHouse schema design: using tenant IDs to partition tables. This can lead to a raft of problems ranging from slow replication to insert errors. The cure is usually to partition by time and sort the data in parts by tenant ID. Problem solved!
- Finally, Alexey Milovidov of ClickHouse Inc gave a talk on “building for fast” that covered the philosophy of building performant systems. Alexey’s talks on performance always contain gems about how to frame performance problems and then solve them. This was no exception. I liked his comment that you always have to dig into details. Don’t just look at averages. Look at why even single requests are slow, and don’t stop until you understand the cause.
The best part of any meetup is the informal meetings before and after the scheduled talks. Here are a few tidbits from conversations.
- We have lots of customers and partners in the Netherlands. It was great to meet you in person! Zoom is not the same as real life.
- ClickHouse Keeper is getting increased usage as a replacement for Zookeeper in new ClickHouse clusters. That’s in turn revealing obscure long-tail bugs of the sort that are difficult to find outside of production. The ClickHouse team now has a full-time engineer to help swat them down quickly.
- Kubernetes came up as a topic several times. One of the participants comes from an org that uses Helm Charts extensively and volunteered to provide one for the Altinity Kubernetes Operator for ClickHouse. Go team!
The conversation continued long after the venue closed and the participants wandered off home or to entertainment in the center of Amsterdam. The weather was late-spring-beautiful and we had an opportunity to see some of the sights, including an amazing number of canals and cool-looking bridges. See the pictures at the end of the article.
In summary, the Amsterdam meetup was definitely worth the trip. We would like to thank ClickHouse Inc for sponsoring this event and look forward to many more in-person events in the future. It’s great to see the ClickHouse community together again.
Speaking of community, Altinity will begin sponsoring in-person events in the US starting in July 2022. Watch for announcements!