json output format

[mostolog]

Hi

Current output format seems human-readable-friendly, but hard to process/use on scripts. Have you considered doing a JSON format version of this tool?

Regards

[misterion]

Hi, it seems to be easy to implement this if decides the output structure of JSON report. If you may help me with it, i has add it fast.

[mostolog]

Hi

I don't know if I'll be useful, but here are my two cents. Let's first help me understand the output of: rma -s myredis -b all > _all.txt

Match : 0% 0/1 [00:00<?, ?it/s] Match : 100% 1/1 [00:00<00:00, 1079.34it/s]

If this is a progress shown during execution, it shouldn't be printed on output but on stdout.

Aggregating keys by pattern and type Apply rules

Useful for stdout, but should be omitted in JSON.

Processing keys: 0% 0/1 [00:00<?, ?it/s] Processing List patterns: 0% 0/1 [00:00<?, ?it/s] Processing List patterns: 100% 1/1 [00:01<00:00, 1.66s/it]

Same as first comment.

Server stat	Stat	Value
Total keys in db	1
RedisDB key space overhead	96
Used hash-max-ziplist-value	64
Used set-max-intset-entries	512
Used zset-max-ziplist-value	64
Used list-max-ziplist-entries	512
Used hash-max-ziplist-entries	512
Used zset-max-ziplist-entries	128
Used list-max-ziplist-value	64
Info used_memory_peak_human	146.72M
Info used_memory_lua	38912
Info used_memory_peak	153846712
Info mem_fragmentation_ratio	1.22
Info mem_allocator	jemalloc-3.6.0
Info used_memory_rss	68235264
Info used_memory	55815136
Info used_memory_human	53.23M

Elements arranged alphabetically with some notes/questions about it

{
    "clustered" : false,
    "clusterName" : null,
    "nodes" : [{
        "clusterRole" : null,
        "info" : {
            "mem_allocator" : "jemalloc-3.6.0",
            "mem_fragmentation_ratio" : 1.22,
            "used_memory" : 55815136,
            "used_memory_human" : "53.23 MB",
            "used_memory_lua" : 38912,
            "used_memory_peak" : 153846712,
            "used_memory_peak_human" : "146.72 MB",
            "used_memory_rss" : 68235264
        },
        "name" : "myredis",
        "used" : {
            "hash-max-ziplist-entries" : 512,
            "hash-max-ziplist-value" : 64,
            "list-max-ziplist-entries" : 512,
            "list-max-ziplist-value" : 64,
            "set-max-intset-entries" : 512,
            "zset-max-ziplist-entries" : 128,
            "zset-max-ziplist-value" : 64
        },
        "redisOverhead" : 96,
        "totalKeys" : 1
    }]
}

• Nodes can have only one cluster role within a cluster, isn't it?
• Is used_memory_human properly calculated? 55815136 Bytes = 55,82 MB! Same for used_memory_peak_human.
• It makes sense to add used_memory_lua_human and used_memory_rss_human fields?

As I'm not an expert on Redis, fields like redisOverhead or zset-max-ziplist-value are totally unknown for me. Even the difference between info and used objects.

Keys by types

Omit.

Match	Count	Type	%
mylist:mytype	1	list	100.00%

[
    {
      "name": "mylist:mytype", "type":"list", "usage": "100"
    }
]

• What's the count column?

Processing list key stats

Guess what...?

Match	Count	Useful	Real	Ratio	Encoding	Min	Max	Avg
mylist:mytype	1	10	64	6.40	embstr [100.0%]	10	10	10
Total:	1	10	64	0.00		0	0	0

{
  "lists" : {
    "name" : "mylist:mytype",
    ...
  },
  "total" : {
    "count" : 1,
    ...
  }
}

Could you elaborate on those columns? are those "usage in bytes"? items in queue?

List stat

What's the difference between keys stats and list stats? AFAIK, i just have 1 key and it's both enlisted on keys stats and list stat

| Match | Count | Avg Count | Min Count | Max Count | Stdev Count | Value mem | Real | Ratio | System | Encoding | Total |

|:-----------|--------:|------------:|------------:|------------:|--------------:|------------:|---------:|--------:|---------:|:--------------------|---------:| | mylist:mytype | 1 | 123021 | 123021 | 123021 | 0 | 42452924 | 48903808 | 1.15 | 3149824 | linkedlist [100.0%] | 52053632 |

| Total: | 1 | 0 | 0 | 0 | 0 | 42452924 | 48903808 | 0.00 | 3149824 | | 52053632 |

Done in 1.58691 seconds

If this information is useful, the result JSON should be divided in 2 objects: the retrieved data and the query metadata:

{
  "request" : { "duration" : 1.58691 },
  "response" : {
    "clustered" : ...
    }
}

[mostolog]

One thing I didn't notice at first, but that IMHO it deserves to be fixed.

Seems the utility is "summarizing" the lists. eg:

myfamily:mylist:mytype1 myfamily:mylist:mytype2

are reported as

myfamily:mylist:*

Again IMHO, it would be much better to have separate entries for each key. Regards

[mostolog]

Thanks!