The concurrency of this library is easier to handle than Sarama.
Notably, it is more compatible with the new model of "almost share
nothing" we use for the inlet and the outlet. The lock for workers in
outlet is removed. We can now use sync.Pool to allocate slice of bytes
in inlet.
It may also be more performant.
In the future, we may want to commit only when pushing data to
ClickHouse. However, this does not seem easy when there is a rebalance.
In case of rebalance, we need to do something when a partition is
revoked to avoid duplicating data. For example, we could flush the
current batch to ClickHouse. Have a look at the
`example/mark_offsets/main.go` file in franz-go repository for a
possible approach. In the meantime, we rely on autocommit.
Another contender could be https://github.com/segmentio/kafka-go. Also
see https://github.com/twmb/franz-go/pull/1064.
There is still room for improvement. For inlet, it would require to know
when Kafka has sent the message (so enabling successes return). For
outlet, it should be possible to reuse the same flow (with a ResetVT
between each use).
This change split the inlet component into a simpler inlet and a new
outlet component. The new inlet component receive flows and put them in
Kafka, unparsed. The outlet component takes them from Kafka and resume
the processing from here (flow parsing, enrichment) and puts them in
ClickHouse.
The main goal is to ensure the inlet does a minimal work to not be late
when processing packets (and restart faster). It also brings some
simplification as the number of knobs to tune everything is reduced: for
inlet, we only need to tune the queue size for UDP, the number of
workers and a few Kafka parameters; for outlet, we need to tune a few
Kafka parameters, the number of workers and a few ClickHouse parameters.
The outlet component features a simple Kafka input component. The core
component becomes just a callback function. There is also a new
ClickHouse component to push data to ClickHouse using the low-level
ch-go library with batch inserts.
This processing has an impact on the internal representation of a
FlowMessage. Previously, it was tailored to dynamically build the
protobuf message to be put in Kafka. Now, it builds the batch request to
be sent to ClickHouse. This makes the FlowMessage structure hides the
content of the next batch request and therefore, it should be reused.
This also changes the way we decode flows as they don't output
FlowMessage anymore, they reuse one that is provided to each worker.
The ClickHouse tables are slightly updated. Instead of using Kafka
engine, the Null engine is used instead.
Fix#1122
Today the timestamp can only be from kernel timetstamp put on the UDP packet
by the kernel.
I propose to add 2 alternative methods of getting the timestamp for netflow/IPFix packets:
- TimestampSourceNetflowPacket: use the timestamp field in the netflow packet itself
- TimestampSourceNetflowFirstSwitched: use the FirstSwitched field from each flow
(the field is actually in uptime, so we need to shift it according to sysUptime)
Using those fields requires the router to have accurate time (probably NTP),
but it allows for architectures where a UDP packet is not immediately
received by the collector, eg. if there is a kafka in-between.
That in turns allows to do maintenance on the collector,
without messing up the statistics
This is a first step to make it accept configuration. Most of the
changes are quite trivial, but I also ran into some difficulties with
query columns and filters. They need the schema for parsing, but parsing
happens before dependencies are instantiated (and even if it was not the
case, parsing is stateless). Therefore, I have added a `Validate()`
method that must be called after instantiation. Various bits `panic()`
if not validated to ensure we catch all cases.
The alternative to make the component manages a global state would have
been simpler but it would break once we add the ability to add or
disable columns.
This is a huge change to make the various subcomponents of the inlet use
the schema to generate the protobuf. For it to make sense, we also
modify the way we parse flows to directly serialize non-essential fields
to Protobuf.
The performance is mostly on par with the previous commit. We are a bit
less efficient because we don't have a fixed structure, but we avoid
loosing too much performance by not relying on reflection and keeping
the production of messages as code. We use less of Goflow2: raw flow
parsing is still done by Goflow2, but we don't use the producer part
anymore. This helps a bit with the performance as we parse less.
Overall, we are 20% than the previous commit and twice faster than the
1.6.4!
```
goos: linux
goarch: amd64
pkg: akvorado/inlet/flow
cpu: AMD Ryzen 5 5600X 6-Core Processor
BenchmarkDecodeEncodeNetflow
BenchmarkDecodeEncodeNetflow/with_encoding
BenchmarkDecodeEncodeNetflow/with_encoding-12 151484 7789 ns/op 8272 B/op 143 allocs/op
BenchmarkDecodeEncodeNetflow/without_encoding
BenchmarkDecodeEncodeNetflow/without_encoding-12 162550 7133 ns/op 8272 B/op 143 allocs/op
BenchmarkDecodeEncodeSflow
BenchmarkDecodeEncodeSflow/with_encoding
BenchmarkDecodeEncodeSflow/with_encoding-12 94844 13193 ns/op 9816 B/op 295 allocs/op
BenchmarkDecodeEncodeSflow/without_encoding
BenchmarkDecodeEncodeSflow/without_encoding-12 92569 12456 ns/op 9816 B/op 295 allocs/op
```
There was a tentative to parse sFlow packets with gopackets, but the
adhoc parser used here is more performant.
This is useful if we cannot tune the sampling rate of the source
equipment and it is too high for us. The sampling rate is adapted.
This is difficult to test, so hopefully, this is correct!
