This repository contains a cli tool to inspect the internal state and log of a Zeebe partition. It is a Java (21) cli application and can be build via maven. It was kicked off during the Camunda Summer Hackdays in 2020 and has been maintained and developed since then.
The following Zeebe versions are supported: [ 8.1, 8.2, 8.3, SNAPSHOT]
When Zeebe is broken there is currently no possibility to find out the last state of Zeebe. If there was no exporter configured or they haven't exported for a while it gets even worse, since it is not clear what the internal engine state is.
To shed some more light in the dark we build a tool called zdb - Zeebe Debugger. It should help you along the way during incidents and broken systems.
Note: To be on the safe side make sure to copy Zeebe data to a separate location, to not mess with a running Zeebe process and mistakingly corrupt any data.
If you have copied data from Zeebe to your local machine you could run the following:
docker run -v <path>/<partitionId>/:/<partitionId>/ ghcr.io/zelldon/zdb log print -p "/<partitionId>"
Important
This approach doesn't work anymore with 8.3+ and non-root containers. See related issue #332
If have Zeebe installed in Kubernetes and want to investigate the Zeebe data you can run zdb as an ephemeral container
kubectl debug -it -c zdb --image=ghcr.io/zelldon/zdb:latest --attach=true --target=zeebe zeebe-0 -- /bin/bash
Alternatively to the strategies above you can download the fat-jar and script and run that locally
cd /usr/bin
curl -O -L https://github.com/Zelldon/zdb/releases/latest/download/zdb
curl -O -L https://github.com/Zelldon/zdb/releases/latest/download/zdb.jar
sed -i 's/target\///' zdb
chmod u+x zdb
zdb --versionUsing zdb you can inspect the internal state or the partition log.
Using zdb you can inspect the internal runtime data or a snapshot.
It shows some information about the current state, incidents, processes, and so on from a single partition.
To inspect the database you should provide the path to the raft-partition/partitions/../runtime/ folder in a partition or one of the snapshot folders raft-partition/partitions/../snapshot/<snapshot-folder>
You then can run several commands to inspect the given state.
Shows the general information of a Zeebe partition. It will show you a statistic (counts) for each existing column family in the state.
zdb state --path=<pathToDatabase>Furthermore, the complete state can be printed as json via the list sub-command.
zdb state list --path=<pathToDatabase>This can be more fine-tuned and a specific column family can be given, such that only key-value pairs are printed to the console.
For example, to see all processes
$ zdb state --path=<pathToDatabase> list -cf PROCESS_CACHE
You can inspect incidents using the following commands.
List all incidents in this partition:
zdb incident list --path=<pathToDatabase>Returns detail to a specific incident:
zdb incident entity <IncidentKey> --path=<pathToDatabase>Find incidents for a given processInstanceKey
zdb incident list --path=<pathToDatabase> | jq '. | map(select(.processInstanceKey==<PI_KEY>))'You can check if there are any processes stuck due to banning using the following commands.
List all banned process instances in this partition:
zdb banned list --path=<pathToDatabase>Returns details to a specific banned instance:
zdb banned entity <ProcessInstanceKey> --path=<pathToDatabase>You can inspect all deployed processes and get the resources of a specific process.
List all deployed processes in this partition:
zdb process list --path=<pathToDatabase>Returns details to a specific process:
zdb process entity <ProcessKey> --path=<pathToDatabase>List all element instances for the given process:
zdb process --path=<pathToDatabase> instances <ProcessKey>You can inspect existing element instances and get details viewed of there state.
Print all information to a given element instance:
zdb instance --path=<pathToDatabase> entity <elementInstanceKey>There is a new (experimental) feature to inspect the state on a generic way with the state subcommand. You can either
print the complete state as json or specify a specific column family (used in Zeebe).
Example to see all processes
$ zdb state --path=<pathToDatabase> list -cf PROCESS_CACHE
You can also inspect the log stream using the command zdb log and his subcommands.
To inspect the log you should provide the path to a specific partition raft-partition/partitions/<partition-id>/.
Shows the general information of a Zeebe partition log, e. g. how many indexes, max. entry size, avg. entry size etc.
zdb log status --path=<pathToPartition>It is possible to inspect the log in more detail and search for a specific index OR position.
To search for a record position use:
zdb log search --path=<pathToPartition> --position=<position>It will print all related information to the record, when it exists in the log.
To search for an index use:
zdb log search --path=<pathToPartition> --index=<position>It will print a details to the specific index, when it exists in the log.
It is possible to print the complete log to standard out. This can be quite helpful if you want to track down some records, which might have caused some issues.
To print the log:
zdb log print --path=<pathToPartition>Per default, the log is printed in JSON format. To pipe it to a file:
zdb log print --path=<pathToPartition> > output.logThe output.log file will contain all records as JSON.
You can limit the printed log via the options --to and --from.
I you want to skip the first records or X positions you can use zdb log print --path=<pathToPartition> --from X whereas X must be a long.
For defining a limit of the print (like until which position the log should be printed) you can use --to like this zdb log print --path=<pathToPartition> --to X whereas X must be a long.
An interesting use case is to print only certain records, for example for only specific process instances.
You can filter the printed log via: --instanceKey
zdb log print -p <pathToPartition> --instanceKey 2251799813686738We support different formats to print the log, like json, table or dot. The json format is used per default. Can be set via -f or --format
Table
zdb log print --format TABLE --path=<pathToPartition>
The table format will print the complete log as space separated table. This can be consumed by other csv tools.
Example:
Index Term Position SourceRecordPosition Timestamp Key RecordType ValueType Intent ProcessInstanceKey BPMNElementType
836304301 304 6888891257 6888891180 1692869671126 2251802375814765 COMMAND PROCESS_INSTANCE ACTIVATE_ELEMENT 2251802375814765 PROCESS
836304301 304 6888891258 6888891180 1692869671126 2251802375814770 EVENT PROCESS_INSTANCE_CREATION CREATED 2251802375814765
836304301 304 6888891259 6888891180 1692869671126 2251802375814765 EVENT PROCESS_INSTANCE ELEMENT_ACTIVATING 2251802375814765 PROCESS
836304301 304 6888891260 6888891180 1692869671126 2251802375814765 EVENT PROCESS_INSTANCE ELEMENT_ACTIVATED 2251802375814765 PROCESS Dot
zdb log print -f dot -p=<pathToPartition>
The dot format will print the complete log as graph in dot language. This can be consumed by graphviz to generate a visual graph of the log.
Generate dot file via:
zdb log print -d -p <pathToPartition> > output.dot
Generate svg:
dot -Tsvg -o test.svg test.dot
Details of a specific column family
zdb state list -p $PATH -cf EXPORTER | jq
{
"data": [
{
"cf": "EXPORTER",
"key": "00 00 00 00 00 00 00 28 00 00 00 0d 65 6c 61 73 74 69 63 73 65 61 72 63 68",
"value": {
"exporterPosition": 619675,
"exporterMetadata": "eyJyZWNvcmRDb3VudGVyc0J5VmFsdWVUeXBlIjp7IkRFUExPWU1FTlQiOjE0LCJQUk9DRVNTX0lOU1RBTkNFIjo1ODcsIklOQ0lERU5UIjo1LCJNRVNTQUdFIjo3OSwiTUVTU0FHRV9TVUJTQ1JJUFRJT04iOjM5LCJQUk9DRVNTX01FU1NBR0VfU1VCU0NSSVBUSU9OIjoxNDksIk1FU1NBR0VfU1RBUlRfRVZFTlRfU1VCU0NSSVBUSU9OIjozNywiVkFSSUFCTEUiOjE5NywiUFJPQ0VTU19JTlNUQU5DRV9DUkVBVElPTiI6MSwiUFJPQ0VTUyI6MTMsIkNPTU1BTkRfRElTVFJJQlVUSU9OIjo4NH19"
}
},
{
"cf": "EXPORTER",
"key": "00 00 00 00 00 00 00 28 00 00 00 0f 4d 65 74 72 69 63 73 45 78 70 6f 72 74 65 72",
"value": {
"exporterPosition": 619676,
"exporterMetadata": ""
}
}
]
}Retrieve a process model
$ zdb state --path=<pathToDatabase> list -cf PROCESS_CACHE | jq --raw-output '.data[0].value.resource' | base64 -d > model.bpmn
zdb comes with autocompletion. Just print it to a file:
zdb generate-completion >> ~/.autocompletions/zdband source that file in your shell profile (i.e. .bash_rc, .zsh_rc, .bash_profile, etc.):
source <(cat $HOME/.autocompletions/zdb)
![dependabot[bot] avatar](https://avatars.githubusercontent.com/in/29110?v=4 "dependabot[bot]")
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent