We currently use the ES 5.5 version and plan to upgrade to the ES 7.4 version.
I deployed two environments,use the same server , 88 core cpu , 256G memory , the cluster have 3 servers, each server starts 3 nodes, the cluster has a total of 9 nodes.
jvm is configured to 30G, others use the default configuration
I used jmeter to write the same data to the cluster,index with 9 shards and 1 replcate. The amount of data is 300G.
When executing the query, the ES 7.4 result took a lot larger than the ES 5.5 , and the ES 7.4 server's io is also very large, there are a lot of disk read conditions.
Query like:
{
"size": 0,
"query": {
"bool": {
"filter": [{
"term": {
"aaa": "bbb"
}
}, {
"range": {
"@timestamp": {
"gte": "1576117714282",
"lte": "1576118614282"
}
}
}
], "Must_not": [{
"terms": {
"bizcode": ["00000000"]
}
}
]
}
},
"aggs": {
"distribution": {
"terms": {
"field": "payer",
"size": 10,
"order": {
"_count": "desc"
}
}
},
"aggs": {
"distribution2": {
"terms": {
"field": "from",
"size": 10,
"order": {
"_count": "desc"
}
}
}
}
}
}
What is the cause of this problem? Are there any changes to the query in ES 7.4?