Description
Elasticsearch version (bin/elasticsearch --version): 7.5 and later
Plugins installed: []
JVM version (java -version):
OS version (uname -a if on a Unix-like system):
Description of the problem including expected versus actual behavior:
Currently when processor used inside foreach modifies field, that foreach is iterating over, it can leads to unexpected results. Especially when it appends anything to the field foreach will iterate over new items as well which leads to infinite loop and it will eventually bring the whole cluster down.
Steps to reproduce:
POST /_ingest/pipeline/_simulate
{
"pipeline": {
"description": "do something",
"processors": [
{
"foreach": {
"field": "names",
"processor": {
"append": {
"field": "names",
"value": "whatever"
}
}
}
}
]
},
"docs": [
{
"_index": "index",
"_id": "id",
"_source": {
"names": [
"anything"
]
}
}
]
}You can invoke it couple of times to make effect faster. Eventually cluster will become unresponsive. On 7.7 and later it should fail pretty fast and nothing will work, on previous versions it will accumulate over time (it took around a day for my cloud instance without any other traffic) and will cause all sorts of slowness (like Kibana loading in 1min or just throwing 500 error)