[CASSANDRA-9744] unrealistic pending compactions displayed

• Type: Bug
• Status: Resolved
• Priority: Normal
• Resolution: Duplicate
• Severity: Normal
• Assignee: Unassigned
• Reporter: Christophe
• Created: 07/Jul/15 04:32
• Updated: 16/Apr/19 09:31
• Resolved: 07/Jul/15 18:12

Description

Hi there,

I am on 2.1.7. I am using DateTieredCompactionStrategy with the following parameters:
min_threshold 16
max_threshold 16
base_time_seconds 300
max_stable_age_days 0.05

Apart from that, nothing exotic. I dump an sstable of about 18M every 12 sec I would say.

Maybe 1 hour after creating my cluster, I can see that:
1. the number of SSTables has constantly been below 60.
2. Most of the time, I have between 1 and 4 pending compaction.
3. I occasionaly have a large number of pending compactions (30-40) for a short period of time (a few min)
4. I witness on 1 node 1610 pending compaction, on another node 367 pending compaction, for a short period of time (a few min)

Item 4 makes me thing there is a bug as I don't see how I can have that many pending compaction with less than 50 sstable.

Issue Links

• duplicates: CASSANDRA-9662 compactionManager reporting wrong pendingtasks