At a glance
The ten commands that appeared most often in the Redis slow log over the last 24 hours, ranked so you can see exactly which operations are stalling the single execution thread. Where SLOWLOG Entries (15m) tells you how many slow commands occurred, this table tells you which ones, so a platform team can go straight to the offending pattern (anO(n)KEYS, an oversizedHGETALL, a slow Lua script) and refactor it.
What it tracks
The card reads Redis’s in-memory slow log (SLOWLOG GET), groups entries by command pattern over the trailing 24 hours, and ranks the top ten by frequency. Each row surfaces the command, how often it crossed the slowlog-log-slower-than threshold (default 10,000 microseconds, or 10 ms), and its typical execution time. Because Redis executes commands on a single thread, every slow command in this list blocked all other clients for the duration shown, so the highest-frequency or highest-duration rows are the most damaging. The usual offenders are full-keyspace scans (KEYS, which should always be replaced with a SCAN cursor loop), O(n) reads on large collections (HGETALL, SMEMBERS, LRANGE 0 -1), unbounded SORT, and slow EVAL/EVALSHA Lua scripts. The 24-hour window is wide enough to catch periodic batch jobs and off-peak cron tasks that a short window would miss. Note that the ranking is bounded by slowlog-max-len (default 128 entries): on a very busy instance older entries are evicted, so a chronically full buffer may under-represent rarer offenders. There is no alert on this card; it is a diagnostic companion to the alerting SLOWLOG count card.
Reconciling against the source
Runredis-cli SLOWLOG GET 128 to dump the raw entries (ID, timestamp, duration in microseconds, command, client) and redis-cli CONFIG GET slowlog-log-slower-than to confirm the threshold the ranking is built against; on ElastiCache, MemoryDB, and Redis Cloud the same SLOWLOG commands work over the standard protocol and the log is also exported to the engine log.