Thanks Jeff for the reply. Answers inlined. Tombstones probably aren't clearing because the same partition exists with older timestamps in other files (this is the "sstableexpiredblockers" problem, or "overlaps"). >>The RF is 2, so there is two copies of one partition in two node. So my >>method to clear expired data doesn't work because of the "overlaps" you >>mentioned. Is my understanding corrent? One more question, nodetool cleanup >>may work for me, but how cleanup deal with the sstable files in TWCS mode? I >>have large sstable files before changing from STCS to TWCS, and newer ones >>with time buckets with TWCS. How does the command deal with them? Compact all >>of them into large sstable files?
If you're certain you are ok losing that data, then you could stop the node, remove lb-143951-big-* , and start the node. This is usually a bad idea in data models that aren't ttl-only time-series, but if you KNOW the data is all expired, and you didnt manually delete any other data, it may work for you. >>My data model is indeed ttl-only time-series. Cheers, -Simon From: Jeff Jirsa Date: 2018-03-27 11:52 To: cassandra Subject: Re: Single sstable file compaction issue Tombstones probably aren't clearing because the same partition exists with older timestamps in other files (this is the "sstableexpiredblockers" problem, or "overlaps"). If you're certain you are ok losing that data, then you could stop the node, remove lb-143951-big-* , and start the node. This is usually a bad idea in data models that aren't ttl-only time-series, but if you KNOW the data is all expired, and you didnt manually delete any other data, it may work for you. On Mon, Mar 26, 2018 at 8:03 PM, wxn...@zjqunshuo.com <wxn...@zjqunshuo.com> wrote: Hi All, I changed STCS to TWCS months ago and left some old sstable files. Some are almost tombstones. To release disk space, I issued compaction command on one file by JMX. After the compaction is done, I got one new file with almost the same size of the old one. Seems no tombstones are cleaned during the compaction. Before compaciton: Max: 01/12/2017 Min: 11/25/2016 Estimated droppable tombstones: 0.9115440366604225 53G Jan 16 00:36 lb-124337-big-Data.db After compaction: Max: 01/12/2017 Min: 11/25/2016 Estimated droppable tombstones: 0.9114708007586322 53G Mar 27 00:17 lb-143951-big-Data.db Questions: 1. Why the compaction didn't clean the tombstones? 2. If one file are all tombstones and I want to it manually(including data, index, filter, etc), do I need to shutdown the node? Cheers, -Simon
