RE: Tombstone warnings in log file
Simon, It means that in processing your queries, Cassandra is going through that many tombstone cells in order to return your results. It is because some of the partitions that you are querying for have already expired. The warning is just cassandra's way of letting you know that your reads are less efficient because you are reading a lot of expired data. You could tune this by altering your tombstone parameters in Cassandra yaml file. But a better solution would be to reduce your GC grace seconds for that table to a smaller value (as opposed to default of 10 days) so that the TTLed data will be purged sooner. You could also consider drafting more efficient queries which won't hit TTLed partitions. Thanks, Meg From: wxn...@zjqunshuo.com [mailto:wxn...@zjqunshuo.com] Sent: Sunday, December 03, 2017 7:49 PM To: user Subject: Tombstone warnings in log file Hi, My cluster is running 2.2.8, no update and deletion, only insertion with TTL. I saw below warnings reacently. What's the meaning of them and what's the impact? WARN [SharedPool-Worker-2] 2017-12-04 09:32:48,833 SliceQueryFilter.java:308 - Read 2461 live and 1978 tombstone cells in cargts.eventdata for key: 129762:20171202 (see tombstone_warn_threshold). 5000 columns were requested, slices=[-] Best regards, -Simon WARN [SharedPool-Worker-2] 2017-12-04 09:32:48,833 SliceQueryFilter.java:308 - Read 2461 live and 1978 tombstone cells in cargts.eventdata for key: 129762:20171202 (see tombstone_warn_threshold). 5000 columns were requested, slices=[-]
RE: Tombstone warnings in log file
Got it. Thank you. From: Meg Mara Date: 2017-12-05 01:54 To: user@cassandra.apache.org Subject: RE: Tombstone warnings in log file Simon, It means that in processing your queries, Cassandra is going through that many tombstone cells in order to return your results. It is because some of the partitions that you are querying for have already expired. The warning is just cassandra’s way of letting you know that your reads are less efficient because you are reading a lot of expired data. You could tune this by altering your tombstone parameters in Cassandra yaml file. But a better solution would be to reduce your GC grace seconds for that table to a smaller value (as opposed to default of 10 days) so that the TTLed data will be purged sooner. You could also consider drafting more efficient queries which won’t hit TTLed partitions. Thanks, Meg From: wxn...@zjqunshuo.com [mailto:wxn...@zjqunshuo.com] Sent: Sunday, December 03, 2017 7:49 PM To: user Subject: Tombstone warnings in log file Hi, My cluster is running 2.2.8, no update and deletion, only insertion with TTL. I saw below warnings reacently. What's the meaning of them and what's the impact? WARN [SharedPool-Worker-2] 2017-12-04 09:32:48,833 SliceQueryFilter.java:308 - Read 2461 live and 1978 tombstone cells in cargts.eventdata for key: 129762:20171202 (see tombstone_warn_threshold). 5000 columns were requested, slices=[-] Best regards, -Simon WARN [SharedPool-Worker-2] 2017-12-04 09:32:48,833 SliceQueryFilter.java:308 - Read 2461 live and 1978 tombstone cells in cargts.eventdata for key: 129762:20171202 (see tombstone_warn_threshold). 5000 columns were requested, slices=[-]
Re: Tombstone warnings in log file
Hello Simon. Tombstone is a tricky topic in Cassandra that brought a lot of questions over time. I exposed my understanding in a blog post last year and thought it might be of interest for you, even though things probably evolved a bit, principles and tuning did not change that much I guess. Here is the post: thelastpickle.com/blog/2016/07/27/about-deletes-and-tombstones.html. 2017-12-05 2:01 GMT+00:00 wxn...@zjqunshuo.com : > Got it. Thank you. > > > *From:* Meg Mara > *Date:* 2017-12-05 01:54 > *To:* user@cassandra.apache.org > *Subject:* RE: Tombstone warnings in log file > > Simon, > > > > It means that in processing your queries, Cassandra is going through that > many tombstone cells in order to return your results. It is because some of > the partitions that you are querying for have already expired. The warning > is just cassandra’s way of letting you know that your reads are less > efficient because you are reading a lot of expired data. > > > > You could tune this by altering your tombstone parameters in Cassandra > yaml file. But a better solution would be to reduce your GC grace seconds > for that table to a smaller value (as opposed to default of 10 days) so > that the TTLed data will be purged sooner. > > > > You could also consider drafting more efficient queries which won’t hit > TTLed partitions. > > > > Thanks, > > *Meg* > > > > > > *From:* wxn...@zjqunshuo.com [mailto:wxn...@zjqunshuo.com] > *Sent:* Sunday, December 03, 2017 7:49 PM > *To:* user > *Subject:* Tombstone warnings in log file > > > > Hi, > > My cluster is running 2.2.8, no update and deletion, only insertion with > TTL. I saw below warnings reacently. What's the meaning of them and what's > the impact? > > > > WARN [SharedPool-Worker-2] 2017-12-04 09:32:48,833 > SliceQueryFilter.java:308 - Read 2461 live and 1978 > tombstone cells in cargts.eventdata for key: 129762:20171202 (see > tombstone_warn_threshold). 5000 columns were requested, slices=[-] > > > > Best regards, > > -Simon > > WARN [SharedPool-Worker-2] 2017-12-04 09:32:48,833 SliceQueryFilter.java:308 > - Read 2461 live and 1978 tombstone cells in cargts.eventdata for > key: 129762:20171202 (see tombstone_warn_threshold). 5000 columns were > requested, slices=[-] > >