Question about 'duplicate' columns
I've been thinking through some cases that I can see happening at some point and thought I'd ask on the list to see if my understanding is correct. Say a bunch of columns have been loaded 'a long time ago', i.e long enough in the past that they have been compacted. My understanding is that if some these columns get reloaded then they are likely to sit in additional sstables until the larger sstable is called up for compaction, which might be a while. The case that springs to mind is filling small gaps in data by doing bulk loads around the gap to make sure that the gap is filled. Have I understood correctly ? thanks -- *Franc Carter* | Systems architect | Sirca Ltd marc.zianideferra...@sirca.org.au franc.car...@sirca.org.au | www.sirca.org.au Tel: +61 2 8355 2514 Level 4, 55 Harrington St, The Rocks NSW 2000 PO Box H58, Australia Square, Sydney NSW 1215
Re: Any good GUI based tool to manage data in Casandra?
There is a list here. http://wiki.apache.org/cassandra/Administration%20Tools Cheers - Aaron Morton Cassandra Consultant New Zealand @aaronmorton http://www.thelastpickle.com On 3/08/2013, at 6:19 AM, Tony Anecito adanec...@yahoo.com wrote: Hi All, Is there a GUI tool for managing data in Cassandra database? I have google and seen tools but they seem to be schema management or explorer to just view data. IT would be great to delete/inset rows or update values for a column via GUI. Thanks, -Tony
Re: cassandra 1.2.5- virtual nodes (num_token) pros/cons?
The reason for me looking at virtual nodes is because of terrible experiences we had with 0.8 repairs and as per documentation (an logically) the virtual nodes seems like it will help repairs being smoother. Is this true? I've not thought too much about how they help repair run smoother, what was the documentation you read ? Also how to get the right number of virtual nodes? Use the default 256 Hope that helps. - Aaron Morton Cassandra Consultant New Zealand @aaronmorton http://www.thelastpickle.com On 3/08/2013, at 7:39 AM, rash aroskar rashmi.aros...@gmail.com wrote: Thanks for helpful responses. The upgrade from 0.8 to 1.2 is not direct, we have setup test cluster where we did upgrade from 0.8 to 1.1 and then 1.2. Also we will do a whole different cluster with 1.2, the 0.8 cluster will not be upgraded. But the data will be moved from 0.8 cluster to 1.2 cluster. The reason for me looking at virtual nodes is because of terrible experiences we had with 0.8 repairs and as per documentation (an logically) the virtual nodes seems like it will help repairs being smoother. Is this true? Also how to get the right number of virtual nodes? David suggested 64 vnodes for 20 machines. Is there a formula or a thought process to be followed to get this number right? On Mon, Jul 29, 2013 at 4:15 AM, aaron morton aa...@thelastpickle.com wrote: I would *strongly* recommend against upgrading from 0.8 directly to 1.2. Skipping a major version is generally not recommended, skipped 3 would seem like carelessness. I second Romain, do the upgrade and make sure the health is good first. +1 but I would also recommend deciding if you actually need to use virtual nodes. The shuffle process can take a long time and people have had mixed experiences with it. If you wanted to move to 1.2 and get vNodes I would consider spinning up a new cluster and bulk loading into it. You could do an initial load and then to delta loads using snapshots, there would however be a period of stale data in the new cluster until the last delta snapshot is loaded. Cheers - Aaron Morton Cassandra Consultant New Zealand @aaronmorton http://www.thelastpickle.com On 27/07/2013, at 3:36 AM, David McNelis dmcne...@gmail.com wrote: I second Romain, do the upgrade and make sure the health is good first. If you have or plan to have a large number of nodes, you might consider using fewer than 256 as your initial vnodes amount. I think that number is inflated from reasonable in the docs, as we've had some people talk about potential performance degradation if you have a large number of nodes and a very high number of vnodes, if I had it to do over again, I'd have done 64 vnodes as my default (across 20 nodes). Another thing to be very cognizant of before shuffle is disk space. You *must* have less than 50% used in order to do the shuffle successfully because no data is removed (cleaned) from a node during the shuffle process and the shuffle process essentially doubles the amount of data until you're able to run a clean. On Fri, Jul 26, 2013 at 11:25 AM, Romain HARDOUIN romain.hardo...@urssaf.fr wrote: Vnodes are a great feature. More nodes are involved during operations such as bootstrap, decommission, etc. DataStax documentation is definitely a must read. That said, If I were you, I'd wait somewhat before to shuffle the ring. I'd focus on cluster upgrade and monitoring the nodes. (number of files handles, memory usage, latency, etc). Upgrading from 0.8 to 1.2 can be tricky, there are so many changes since then. Be careful about compaction strategies you choose and double check the options. Regards, Romain rash aroskar rashmi.aros...@gmail.com a écrit sur 25/07/2013 23:25:11 : De : rash aroskar rashmi.aros...@gmail.com A : user@cassandra.apache.org, Date : 25/07/2013 23:25 Objet : cassandra 1.2.5- virtual nodes (num_token) pros/cons? Hi, I am upgrading my cassandra cluster from 0.8 to 1.2.5. In cassandra 1.2.5 the 'num_token' attribute confuses me. I understand that it distributes multiple tokens per node but I am not clear how that is helpful for performance or load balancing. Can anyone elaborate? has anyone used this feature and knows its advantages/disadvantages? Thanks, Rashmi
Re: Better to have lower or greater cardinality for partition key in CQL3?
So from anyones experience, is it better to use a low cardinality partition key or a high cardinality. IMHO go with whatever best supports the read paths. They all get If you have lots (e.g. north of 1 billion) rows per node there are extra considerations that come into play. Cassandra 1.2 helps a lot with the bloom filters and compression meta off heap. Basically you may need to pay more attention to memory usage at that scale. This is one place for the LCS can help. It allows you to have a higher bloom filter FP chance, which results in a lower memory overhead for a given number of rows. Remember that LCS uses roughly twice the IO though, so make sure you can handle the throughput. Otherwise your update workflow sounds is a perfect match for Size Tiered compaction. Hope that helps. - Aaron Morton Cassandra Consultant New Zealand @aaronmorton http://www.thelastpickle.com On 5/08/2013, at 4:57 PM, David Ward da...@shareablee.com wrote: Hello, Was curious what people had found to be better for structuring/modeling data into C*? With my data I have two primary keys, one 64 bit int thats 0 - 50 million ( its unlikely to go higher then 70 million ever ) and another 64 bit that's probably close to hitting a trillion in the next year or so. Looking at how the data is going to behave, for the first few months each row/record will be updated but after that its practically written in stone. Still I was leaning toward leveled compaction as it gets updated anywhere from once an hour to at least once a day for the first 7 days. So from anyones experience, is it better to use a low cardinality partition key or a high cardinality. Additionally data organized by the low cardinality set is probably 1-6B ( and growing ) but the high cardinality would be 1-6MB only 2-3x a year. Thanks, Dave new high cardinality keys in 1 year ~15,768,00,000 new low cardinality keys in 1 year = 10,000-30,000 low cardinality key set size ~1-6GB high cardinality key set size 1-5MB
Re: Which of these VPS configurations would perform better for Cassandra ?
how many nodes to start with(2 ok?) ? I'd recommend 3, that will give you some redundancy see http://thelastpickle.com/2011/06/13/Down-For-Me/ Cheers - Aaron Morton Cassandra Consultant New Zealand @aaronmorton http://www.thelastpickle.com On 5/08/2013, at 1:41 AM, Rajkumar Gupta rajkumar@gmail.com wrote: okay, so what should a workable VPS configuration to start with minimum how many nodes to start with(2 ok?) ? Seriously I cannot afford the tensions of colocation setup. My hosting provider provides SSD drives with KVM virtualization.
Re: Reducing the number of vnodes
Repair runs in two phases, first it works out the differences then it streams the data. The length of the first depends on the size of the data and the second on the level of inconsistency. To track the first use nodetool compaction stats or look in the logs for the messages about requesting or receiving Merkle Tree's. To track the second use nodetool netstats or look in the logs for messages about streaming X number of ranges. Hope that helps. - Aaron Morton Cassandra Consultant New Zealand @aaronmorton http://www.thelastpickle.com On 6/08/2013, at 1:41 AM, Christopher Wirt chris.w...@struq.com wrote: 1.2.4. Really hesitant to upgrade versions due to the inevitable issues it will cause. Guess I could upgrade a single node and let that run for a while before upgrading all nodes. From: Haithem Jarraya [mailto:a-hjarr...@expedia.com] Sent: 05 August 2013 13:04 To: user@cassandra.apache.org Subject: Re: Reducing the number of vnodes Chris, Which C* version are you running? You might want to do an upgrade to the latest version before reducing the vnode counts, a lot of fixes and improvement went in lately, it might help you getting your repair faster. H On 5 Aug 2013, at 12:30, Christopher Wirt chris.w...@struq.com wrote: Hi, I’m thinking about reducing the number of vnodes per server. We have 3 DC setup – one with 9 nodes, two with 3 nodes each. Each node has 256 vnodes. We’ve found that repair operations are beginning to take too long. Is reducing the number of vnodes to 64/32 likely to help our situation? What options do I have for achieving this in a live cluster? Thanks, Chris
Re: Unable to bootstrap node
Caused by: java.io.FileNotFoundException: /data/1/cassandra/data/rts/40301_feedProducts/rts-40301_feedProducts-ib-1-Data.db (No such file or directory) at java.io.RandomAccessFile.open(Native Method) at java.io.RandomAccessFile.init(RandomAccessFile.java:233) at org.apache.cassandra.io.util.RandomAccessReader.init(RandomAccessReader.java:67) at org.apache.cassandra.io.compress.CompressedRandomAccessReader.init This is somewhat serous, specially if it's from the a bug in dropping tables. Though I would expect that would show up for a lot of people. Does the file exist on disk? Are the permissions correct ? IMHO you need to address this issue on the existing nodes before worrying about the new node. Cheers - Aaron Morton Cassandra Consultant New Zealand @aaronmorton http://www.thelastpickle.com On 6/08/2013, at 1:25 PM, sankalp kohli kohlisank...@gmail.com wrote: Let me know if this fixes the problem? On Mon, Aug 5, 2013 at 6:24 PM, sankalp kohli kohlisank...@gmail.com wrote: So the problem is that when you dropped and recreated the table with the same name, some how the old CFStore object was not purged. So now there were two objects which caused same sstable to have 2 SSTableReader object. The fix is to find all nodes which is emitting this FileNotFound Exception and restart them. In your case, restart the node which is serving the data and emitting FileNotFound exception. Once this is up, again restart the bootstrapping node with bootstrap argument. Now it will successfully stream the data. On Mon, Aug 5, 2013 at 6:08 PM, Keith Wright kwri...@nanigans.com wrote: Yes we likely dropped and recreated tables. If we stop the sending node, what will happen to the bootstrapping node? sankalp kohli kohlisank...@gmail.com wrote: Hi, The problem is that the node sending the stream is hitting this FileNotFound exception. You need to restart this node and it should fix the problem. Are you seeing lot of FileNotFoundExceptions? Did you do any schema change recently? Sankalp On Mon, Aug 5, 2013 at 5:39 PM, Keith Wright kwri...@nanigans.com wrote: Hi all, I have been trying to bootstrap a new node into my 7 node 1.2.4 C* cluster with Vnodes RF3 with no luck. It gets close to completing and then the streaming just stalls with streaming at 99% from 1 or 2 nodes. Nodetool netstats shows the items that have yet to stream but the logs on the new node do not show any errors. I tried shutting down then node, clearing all data/commit logs/caches, and re-boot strapping with no luck. The nodes that are hanging sending the data only have the error below but that's related to compactions (see below) although it is one of the files that is waiting to be sent. I tried nodetool scrub on the column family with the missing item but got an error indicating it could not get a hard link. Any ideas? We were able to bootstrap one of the new nodes with no issues but this other one has been a real pain. Note that when the new node is joining the cluster, it does not appear in nodetool status. Is that expected? Thanks all, my next step is to try getting a new IP for this machine, my thought being that the cluster doesn't like me continuing to attempt to bootstrap the node repeatedly each time getting a new host id. [kwright@lxpcas008 ~]$ nodetool netstats | grep rts-40301_feedProducts-ib-1-Data.db rts: /data/1/cassandra/data/rts/40301_feedProducts/rts-40301_feedProducts-ib-1-Data.db sections=73 progress=0/1884669 - 0% ERROR [ReadStage:427] 2013-08-05 23:23:29,294 CassandraDaemon.java (line 174) Exception in thread Thread[ReadStage:427,5,main] java.lang.RuntimeException: java.io.FileNotFoundException: /data/1/cassandra/data/rts/40301_feedProducts/rts-40301_feedProducts-ib-1-Data.db (No such file or directory) at org.apache.cassandra.io.compress.CompressedRandomAccessReader.open(CompressedRandomAccessReader.java:46) at org.apache.cassandra.io.util.CompressedSegmentedFile.createReader(CompressedSegmentedFile.java:57) at org.apache.cassandra.io.util.PoolingSegmentedFile.getSegment(PoolingSegmentedFile.java:41) at org.apache.cassandra.io.sstable.SSTableReader.getFileDataInput(SSTableReader.java:976) at org.apache.cassandra.db.columniterator.SSTableNamesIterator.createFileDataInput(SSTableNamesIterator.java:98) at org.apache.cassandra.db.columniterator.SSTableNamesIterator.read(SSTableNamesIterator.java:117) at org.apache.cassandra.db.columniterator.SSTableNamesIterator.init(SSTableNamesIterator.java:64) at org.apache.cassandra.db.filter.NamesQueryFilter.getSSTableColumnIterator(NamesQueryFilter.java:81) at org.apache.cassandra.db.filter.QueryFilter.getSSTableColumnIterator(QueryFilter.java:68) at
Re: Question about 'duplicate' columns
Yes. If you overwrite much older data with new data both versions of the column will remain on disk until compaction get's to work on both fragments of the row. Cheers - Aaron Morton Cassandra Consultant New Zealand @aaronmorton http://www.thelastpickle.com On 6/08/2013, at 6:48 PM, Franc Carter franc.car...@sirca.org.au wrote: I've been thinking through some cases that I can see happening at some point and thought I'd ask on the list to see if my understanding is correct. Say a bunch of columns have been loaded 'a long time ago', i.e long enough in the past that they have been compacted. My understanding is that if some these columns get reloaded then they are likely to sit in additional sstables until the larger sstable is called up for compaction, which might be a while. The case that springs to mind is filling small gaps in data by doing bulk loads around the gap to make sure that the gap is filled. Have I understood correctly ? thanks -- Franc Carter | Systems architect | Sirca Ltd franc.car...@sirca.org.au | www.sirca.org.au Tel: +61 2 8355 2514 Level 4, 55 Harrington St, The Rocks NSW 2000 PO Box H58, Australia Square, Sydney NSW 1215
Re: Question about 'duplicate' columns
On Tue, Aug 6, 2013 at 6:10 PM, Aaron Morton aa...@thelastpickle.comwrote: Yes. If you overwrite much older data with new data both versions of the column will remain on disk until compaction get's to work on both fragments of the row. thanks Cheers - Aaron Morton Cassandra Consultant New Zealand @aaronmorton http://www.thelastpickle.com On 6/08/2013, at 6:48 PM, Franc Carter franc.car...@sirca.org.au wrote: I've been thinking through some cases that I can see happening at some point and thought I'd ask on the list to see if my understanding is correct. Say a bunch of columns have been loaded 'a long time ago', i.e long enough in the past that they have been compacted. My understanding is that if some these columns get reloaded then they are likely to sit in additional sstables until the larger sstable is called up for compaction, which might be a while. The case that springs to mind is filling small gaps in data by doing bulk loads around the gap to make sure that the gap is filled. Have I understood correctly ? thanks -- *Franc Carter* | Systems architect | Sirca Ltd marc.zianideferra...@sirca.org.au franc.car...@sirca.org.au | www.sirca.org.au Tel: +61 2 8355 2514 Level 4, 55 Harrington St, The Rocks NSW 2000 PO Box H58, Australia Square, Sydney NSW 1215 -- *Franc Carter* | Systems architect | Sirca Ltd marc.zianideferra...@sirca.org.au franc.car...@sirca.org.au | www.sirca.org.au Tel: +61 2 8355 2514 Level 4, 55 Harrington St, The Rocks NSW 2000 PO Box H58, Australia Square, Sydney NSW 1215
Is there update-in-place on maps?
Hi, I think it does not fit the model of how C* does writes, but just to verify: Is there an update-in-place possibility on maps? That is, could I do an atomic increment on a value in a map? Jan
Effect of TTL on collection updates
Hi, after seeing Patrick's truly excellent 3-part series on modeling, this question pops up: When I do an update on a collection, using a TTL in the update statement (like Patrick does in the example with the login-location time series example), does the TTL apply to the update only, or to the row as a whole? Jan
Re: cassandra 1.2.5- virtual nodes (num_token) pros/cons?
On 6 August 2013 08:40, Aaron Morton aa...@thelastpickle.com wrote: The reason for me looking at virtual nodes is because of terrible experiences we had with 0.8 repairs and as per documentation (an logically) the virtual nodes seems like it will help repairs being smoother. Is this true? I've not thought too much about how they help repair run smoother, what was the documentation you read ? There might be a slight improvement but I haven't observed any. The difference might be that, because every node shares replicas with every other (with high probability), a single repair operation does the same work on the node it was called on, but the rest is spread out over the cluster, rather than just the RF nodes either side of the repairing node. This means the post-repair compaction work will take less time and the length of time a node is loaded for during repair is less. However, the other benefits of vnodes are likely to be much more useful. Richard.
Re: Effect of TTL on collection updates
Hi Jan TTLs if used only apply to the newly inserted/updated values, from : http://cassandra.apache.org/doc/cql3/CQL.html#collections This manual is updated often enough to be up to date, and so, useful, you should keep it bookmarked. Alain 2013/8/6 Jan Algermissen jan.algermis...@nordsc.com Hi, after seeing Patrick's truly excellent 3-part series on modeling, this question pops up: When I do an update on a collection, using a TTL in the update statement (like Patrick does in the example with the login-location time series example), does the TTL apply to the update only, or to the row as a whole? Jan
Re: Is there update-in-place on maps?
Once again, this should answer your question : http://cassandra.apache.org/doc/cql3/CQL.html#collections Alain 2013/8/6 Jan Algermissen jan.algermis...@nordsc.com Hi, I think it does not fit the model of how C* does writes, but just to verify: Is there an update-in-place possibility on maps? That is, could I do an atomic increment on a value in a map? Jan
Re: Is there update-in-place on maps?
Alain, On 06.08.2013, at 11:17, Alain RODRIGUEZ arodr...@gmail.com wrote: Once again, this should answer your question : http://cassandra.apache.org/doc/cql3/CQL.html#collections yup, I understand the hint :-) However, since I am about to base application architecture on these capabilities, I wanted to make sure I do not read anything into the docs that isn't there. As for the atomic increment, I take the answer is 'no, there is no atomic increment, I have to pull the value to the client and send an update with the new value'. Jan Alain 2013/8/6 Jan Algermissen jan.algermis...@nordsc.com Hi, I think it does not fit the model of how C* does writes, but just to verify: Is there an update-in-place possibility on maps? That is, could I do an atomic increment on a value in a map? Jan
Re: Is there update-in-place on maps?
Store pointers to counters as map values?
Re: Is there update-in-place on maps?
On 06.08.2013, at 11:36, Andy Twigg andy.tw...@gmail.com wrote: Store pointers to counters as map values? Sorry, but this fits into nothing I know about C* so far - can you explain? Jan
Re: Is there update-in-place on maps?
Counters can be atomically incremented ( http://wiki.apache.org/cassandra/Counters). Pick a UUID for the counter, and use that: c=map.get(k); c.incr() On 6 August 2013 11:01, Jan Algermissen jan.algermis...@nordsc.com wrote: On 06.08.2013, at 11:36, Andy Twigg andy.tw...@gmail.com wrote: Store pointers to counters as map values? Sorry, but this fits into nothing I know about C* so far - can you explain? Jan -- Dr Andy Twigg Junior Research Fellow, St Johns College, Oxford Room 351, Department of Computer Science http://www.cs.ox.ac.uk/people/andy.twigg/ andy.tw...@cs.ox.ac.uk | +447799647538
Re: Any good GUI based tool to manage data in Casandra?
Thanks Aaron. I found that before I asked the question and Helenos seems the closest but it does not allow you to easily use CRUD like say SQL Server Management tools where you can get a list of say 1,000 records in a grid control and select rows for deletion or insert or update. I will look closer at that one since this is the reply from the team but if users on this email list have other suggestions please do not hesitate to reply. Many Thanks, -Tony From: Aaron Morton aa...@thelastpickle.com To: Cassandra User user@cassandra.apache.org Sent: Tuesday, August 6, 2013 1:38 AM Subject: Re: Any good GUI based tool to manage data in Casandra? There is a list here. http://wiki.apache.org/cassandra/Administration%20Tools Cheers - Aaron Morton Cassandra Consultant New Zealand @aaronmorton http://www.thelastpickle.com/ On 3/08/2013, at 6:19 AM, Tony Anecito adanec...@yahoo.com wrote: Hi All, Is there a GUI tool for managing data in Cassandra database? I have google and seen tools but they seem to be schema management or explorer to just view data. IT would be great to delete/inset rows or update values for a column via GUI. Thanks, -Tony
clarification of token() in CQL3
I've seen in several places the advice to use queries like to this page through lots of rows: select id from mytable where token(id) token(last_id) But it's hard to find detailed information about how this works (at least that I can understand -- the description in the Cassandra manual is pretty brief). One thing I'd like to know is if new rows are always guaranteed to have token(new_id) token(ids-of-all-previous-rows)? E.g. if I have one process that adds rows to a table, and another that processes rows from the table, can the processor save the id of the last row processed and when he wakes up use: select * from mytable where token(id) token(last_processed_id) to process only new rows? Will this always work to get only new rows?
Re: clarification of token() in CQL3
On 6 August 2013 15:12, Keith Freeman 8fo...@gmail.com wrote: I've seen in several places the advice to use queries like to this page through lots of rows: select id from mytable where token(id) token(last_id) But it's hard to find detailed information about how this works (at least that I can understand -- the description in the Cassandra manual is pretty brief). One thing I'd like to know is if new rows are always guaranteed to have token(new_id) token(ids-of-all-previous-rows)? E.g. if I have one process that adds rows to a table, and another that processes rows from the table, can the processor save the id of the last row processed and when he wakes up use: select * from mytable where token(id) token(last_processed_id) to process only new rows? Will this always work to get only new rows? No, unfortunately not. The tokens are generated by the partitioner - they are the hash of the row key. New tokens could be anywhere in the range of tokens so you can't use token ordering to find new rows. The query you suggest works to page through all the data in your column family. Rows will be returned regardless of when they were added (as long as they were added before the query started). Finding rows that have been added since a certain time is hard in Cassandra since they are stored in token order. In general you have to read through all the data and work out from e.g. a date field if they should be treated as new. Richard.
RE: Counters and replication
Hi Richard, Thanks for your reply. The uid value is a generated guid and should distribute nicely. I've just checked the data yesterday there are only 3 uids out of millions for which there would have been more than 1000 increments. We started with 256 num_tokens. Client and server side I can see the writes being balanced. Anyway, think I've got things under control now. I appears I hadn't set an sstable size on the cf compaction strategy (LCS). I guess this was defaulting to 10MB. After setting this to 256MB one of the 'bad' nodes fixed itself. The other two appeared to stall mid compaction, but after a quick restart both resumed compaction with acceptable CPU utilization. Any insight into how this caused the issue is welcome. Thanks From: Richard Low [mailto:rich...@wentnet.com] Sent: 05 August 2013 20:30 To: user@cassandra.apache.org Subject: Re: Counters and replication On 5 August 2013 20:04, Christopher Wirt chris.w...@struq.com wrote: Hello, Question about counters, replication and the ReplicateOnWriteStage I've recently turned on a new CF which uses a counter column. We have a three DC setup running Cassandra 1.2.4 with vNodes, hex core processors, 32Gb memory. DC 1 - 9 nodes with RF 3 DC 2 - 3 nodes with RF 2 DC 3 - 3 nodes with RF 2 DC 1 one receives most of the updates to this counter column. ~3k per sec. I've disabled any client reads while I sort out this issue. Disk utilization is very low Memory is aplenty (while not reading) Schema: CREATE TABLE cf1 ( uid uuid, id1 int, id2 int, id3 int, ct counter, PRIMARY KEY (uid, id1, id2, id3) ) WITH . Three of the machines in DC 1 are reporting very high CPU load. Looking at tpstats there is a large number of pending ReplicateOnWriteStage just on those machines. Why would only three of the machines be reporting this? Assuming its distributed by uuid value there should be an even load across the cluster, yea? Am I missing something about how distributed counters work? If you have many different uid values and your cluster is balanced then you should see even load. Were your tokens chosen randomly? Did you start out with num_tokens set high or upgrade from num_tokens=1 or an earlier Cassandra version? Is it possible your workload is incrementing the counter for one particular uid much more than the others? The distribution of counters works the same as for non-counters in terms of which nodes receive which values. However, there is a read on the coordinator (randomly chosen for each inc) to read the current value and replicate it to the remaining replicas. This makes counter increments much more expensive than normal inserts, even if all your counters fit in cache. This is done in the ReplicateOnWriteStage, which is why you are seeing that queue build up. Is changing CL to ONE fine if I'm not too worried about 100% consistency? Yes, but to make the biggest difference you will need to turn off replicate_on_write (alter table cf1 with replicate_on_write = false;) but this *guarantees* your counts aren't replicated, even if all replicas are up. It avoids doing the read, so makes a huge difference to performance, but means that if a node is unavailable later on, you *will* read inconsistent counts. (Or, worse, if a node fails, you will lose counts forever.) This is in contrast to CL.ONE inserts for normal values when inserts are still attempted on all replicas, but only one is required to succeed. So you might be able to get a temporary performance boost by changing replicate_on_write if your counter values aren't important. But this won't solve the root of the problem. Richard.
Re: clarification of token() in CQL3
Ok, I get that, I'll have to find another way to sort out new rows. Your description makes me think that if new rows are added during the paging (i.e. between one select with token()'s and another), they might show up in the query results, right? (because the hash of the new row keys might fall sequentially after token(last_processed_row)) On 08/06/2013 08:18 AM, Richard Low wrote: On 6 August 2013 15:12, Keith Freeman 8fo...@gmail.com mailto:8fo...@gmail.com wrote: I've seen in several places the advice to use queries like to this page through lots of rows: select id from mytable where token(id) token(last_id) But it's hard to find detailed information about how this works (at least that I can understand -- the description in the Cassandra manual is pretty brief). One thing I'd like to know is if new rows are always guaranteed to have token(new_id) token(ids-of-all-previous-rows)? E.g. if I have one process that adds rows to a table, and another that processes rows from the table, can the processor save the id of the last row processed and when he wakes up use: select * from mytable where token(id) token(last_processed_id) to process only new rows? Will this always work to get only new rows? No, unfortunately not. The tokens are generated by the partitioner - they are the hash of the row key. New tokens could be anywhere in the range of tokens so you can't use token ordering to find new rows. The query you suggest works to page through all the data in your column family. Rows will be returned regardless of when they were added (as long as they were added before the query started). Finding rows that have been added since a certain time is hard in Cassandra since they are stored in token order. In general you have to read through all the data and work out from e.g. a date field if they should be treated as new. Richard.
Re: clarification of token() in CQL3
On 6 August 2013 16:56, Keith Freeman 8fo...@gmail.com wrote: Your description makes me think that if new rows are added during the paging (i.e. between one select with token()'s and another), they might show up in the query results, right? (because the hash of the new row keys might fall sequentially after token(last_processed_row)) Yes, new rows will appear if their hash is greater than last_processed_row. Richard.
CQL3 select between is broken?
I've been looking at examples about modeling series data in Cassandra, and in one experiment created a table like this: create table vvv (k text, t bigint, value text, primary key (k, t)); After inserting some data with identical k values and differing t values, I tried this query (which is nearly identical to another example I found on the mailing list): cqlsh:smdbxp select * from vvv where k = 'a' and t between 111 and 222; Bad Request: line 1:54 no viable alternative at input '222' Why doesn't that work? Is the syntax of the select wrong for CQL3? (I'm running 1.2.8)
Re: CQL3 select between is broken?
http://cassandra.apache.org/doc/cql3/CQL.html#selectStmt try `and t 111 and t 222' or = and = if you want inclusive. On Tue, Aug 6, 2013 at 10:35 AM, Keith Freeman 8fo...@gmail.com wrote: I've been looking at examples about modeling series data in Cassandra, and in one experiment created a table like this: create table vvv (k text, t bigint, value text, primary key (k, t)); After inserting some data with identical k values and differing t values, I tried this query (which is nearly identical to another example I found on the mailing list): cqlsh:smdbxp select * from vvv where k = 'a' and t between 111 and 222; Bad Request: line 1:54 no viable alternative at input '222' Why doesn't that work? Is the syntax of the select wrong for CQL3? (I'm running 1.2.8)
Re: Unable to bootstrap node
The file does not appear on disk and the permissions are definitely correct. We have seen the file in snapshots. This is completely blocking us from adding the new node. How can we recover? Just run repairs? Thanks From: Aaron Morton aa...@thelastpickle.commailto:aa...@thelastpickle.com Reply-To: user@cassandra.apache.orgmailto:user@cassandra.apache.org user@cassandra.apache.orgmailto:user@cassandra.apache.org Date: Tuesday, August 6, 2013 4:06 AM To: user@cassandra.apache.orgmailto:user@cassandra.apache.org user@cassandra.apache.orgmailto:user@cassandra.apache.org Subject: Re: Unable to bootstrap node Caused by: java.io.FileNotFoundException: /data/1/cassandra/data/rts/40301_feedProducts/rts-40301_feedProducts-ib-1-Data.db (No such file or directory) at java.io.RandomAccessFile.open(Native Method) at java.io.RandomAccessFile.init(RandomAccessFile.java:233) at org.apache.cassandra.io.util.RandomAccessReader.init(RandomAccessReader.java:67) at org.apache.cassandra.io.compress.CompressedRandomAccessReader.init This is somewhat serous, specially if it's from the a bug in dropping tables. Though I would expect that would show up for a lot of people. Does the file exist on disk? Are the permissions correct ? IMHO you need to address this issue on the existing nodes before worrying about the new node. Cheers - Aaron Morton Cassandra Consultant New Zealand @aaronmorton http://www.thelastpickle.com On 6/08/2013, at 1:25 PM, sankalp kohli kohlisank...@gmail.commailto:kohlisank...@gmail.com wrote: Let me know if this fixes the problem? On Mon, Aug 5, 2013 at 6:24 PM, sankalp kohli kohlisank...@gmail.commailto:kohlisank...@gmail.com wrote: So the problem is that when you dropped and recreated the table with the same name, some how the old CFStore object was not purged. So now there were two objects which caused same sstable to have 2 SSTableReader object. The fix is to find all nodes which is emitting this FileNotFound Exception and restart them. In your case, restart the node which is serving the data and emitting FileNotFound exception. Once this is up, again restart the bootstrapping node with bootstrap argument. Now it will successfully stream the data. On Mon, Aug 5, 2013 at 6:08 PM, Keith Wright kwri...@nanigans.commailto:kwri...@nanigans.com wrote: Yes we likely dropped and recreated tables. If we stop the sending node, what will happen to the bootstrapping node? sankalp kohli kohlisank...@gmail.commailto:kohlisank...@gmail.com wrote: Hi, The problem is that the node sending the stream is hitting this FileNotFound exception. You need to restart this node and it should fix the problem. Are you seeing lot of FileNotFoundExceptions? Did you do any schema change recently? Sankalp On Mon, Aug 5, 2013 at 5:39 PM, Keith Wright kwri...@nanigans.commailto:kwri...@nanigans.com wrote: Hi all, I have been trying to bootstrap a new node into my 7 node 1.2.4 C* cluster with Vnodes RF3 with no luck. It gets close to completing and then the streaming just stalls with streaming at 99% from 1 or 2 nodes. Nodetool netstats shows the items that have yet to stream but the logs on the new node do not show any errors. I tried shutting down then node, clearing all data/commit logs/caches, and re-boot strapping with no luck. The nodes that are hanging sending the data only have the error below but that's related to compactions (see below) although it is one of the files that is waiting to be sent. I tried nodetool scrub on the column family with the missing item but got an error indicating it could not get a hard link. Any ideas? We were able to bootstrap one of the new nodes with no issues but this other one has been a real pain. Note that when the new node is joining the cluster, it does not appear in nodetool status. Is that expected? Thanks all, my next step is to try getting a new IP for this machine, my thought being that the cluster doesn't like me continuing to attempt to bootstrap the node repeatedly each time getting a new host id. [kwright@lxpcas008 ~]$ nodetool netstats | grep rts-40301_feedProducts-ib-1-Data.db rts: /data/1/cassandra/data/rts/40301_feedProducts/rts-40301_feedProducts-ib-1-Data.db sections=73 progress=0/1884669 - 0% ERROR [ReadStage:427] 2013-08-05 23:23:29,294 CassandraDaemon.java (line 174) Exception in thread Thread[ReadStage:427,5,main] java.lang.RuntimeException: java.io.FileNotFoundException: /data/1/cassandra/data/rts/40301_feedProducts/rts-40301_feedProducts-ib-1-Data.db (No such file or directory) at org.apache.cassandra.io.compress.CompressedRandomAccessReader.open(CompressedRandomAccessReader.java:46) at org.apache.cassandra.io.util.CompressedSegmentedFile.createReader(CompressedSegmentedFile.java:57) at org.apache.cassandra.io.util.PoolingSegmentedFile.getSegment(PoolingSegmentedFile.java:41) at
Re: Unable to bootstrap node
@Aaron This problem happens when you drop and recreate a keyspace with the same name and you do it very quickly. I have also filed a JIRA for it https://issues.apache.org/jira/browse/CASSANDRA-5843 On Tue, Aug 6, 2013 at 10:31 AM, Keith Wright kwri...@nanigans.com wrote: The file does not appear on disk and the permissions are definitely correct. We have seen the file in snapshots. This is completely blocking us from adding the new node. How can we recover? Just run repairs? Thanks From: Aaron Morton aa...@thelastpickle.com Reply-To: user@cassandra.apache.org user@cassandra.apache.org Date: Tuesday, August 6, 2013 4:06 AM To: user@cassandra.apache.org user@cassandra.apache.org Subject: Re: Unable to bootstrap node Caused by: java.io.FileNotFoundException: /data/1/cassandra/data/rts/40301_feedProducts/rts-40301_feedProducts-ib-1-Data.db (No such file or directory) at java.io.RandomAccessFile.open(Native Method) at java.io.RandomAccessFile.init(RandomAccessFile.java:233) at org.apache.cassandra.io.util.RandomAccessReader.init(RandomAccessReader.java:67) at org.apache.cassandra.io.compress.CompressedRandomAccessReader.init This is somewhat serous, specially if it's from the a bug in dropping tables. Though I would expect that would show up for a lot of people. Does the file exist on disk? Are the permissions correct ? IMHO you need to address this issue on the existing nodes before worrying about the new node. Cheers - Aaron Morton Cassandra Consultant New Zealand @aaronmorton http://www.thelastpickle.com On 6/08/2013, at 1:25 PM, sankalp kohli kohlisank...@gmail.com wrote: Let me know if this fixes the problem? On Mon, Aug 5, 2013 at 6:24 PM, sankalp kohli kohlisank...@gmail.comwrote: So the problem is that when you dropped and recreated the table with the same name, some how the old CFStore object was not purged. So now there were two objects which caused same sstable to have 2 SSTableReader object. The fix is to find all nodes which is emitting this FileNotFound Exception and restart them. In your case, restart the node which is serving the data and emitting FileNotFound exception. Once this is up, again restart the bootstrapping node with bootstrap argument. Now it will successfully stream the data. On Mon, Aug 5, 2013 at 6:08 PM, Keith Wright kwri...@nanigans.comwrote: Yes we likely dropped and recreated tables. If we stop the sending node, what will happen to the bootstrapping node? sankalp kohli kohlisank...@gmail.com wrote: Hi, The problem is that the node sending the stream is hitting this FileNotFound exception. You need to restart this node and it should fix the problem. Are you seeing lot of FileNotFoundExceptions? Did you do any schema change recently? Sankalp On Mon, Aug 5, 2013 at 5:39 PM, Keith Wright kwri...@nanigans.comwrote: Hi all, I have been trying to bootstrap a new node into my 7 node 1.2.4 C* cluster with Vnodes RF3 with no luck. It gets close to completing and then the streaming just stalls with streaming at 99% from 1 or 2 nodes. Nodetool netstats shows the items that have yet to stream but the logs on the new node do not show any errors. I tried shutting down then node, clearing all data/commit logs/caches, and re-boot strapping with no luck. The nodes that are hanging sending the data only have the error below but that's related to compactions (see below) although it is one of the files that is waiting to be sent. I tried nodetool scrub on the column family with the missing item but got an error indicating it could not get a hard link. Any ideas? We were able to bootstrap one of the new nodes with no issues but this other one has been a real pain. Note that when the new node is joining the cluster, it does not appear in nodetool status. Is that expected? Thanks all, my next step is to try getting a new IP for this machine, my thought being that the cluster doesn't like me continuing to attempt to bootstrap the node repeatedly each time getting a new host id. [kwright@lxpcas008 ~]$ nodetool netstats | grep rts-40301_feedProducts-ib-1-Data.db rts: /data/1/cassandra/data/rts/40301_feedProducts/rts-40301_feedProducts-ib-1-Data.db sections=73 progress=0/1884669 - 0% ERROR [ReadStage:427] 2013-08-05 23:23:29,294 CassandraDaemon.java (line 174) Exception in thread Thread[ReadStage:427,5,main] java.lang.RuntimeException: java.io.FileNotFoundException: /data/1/cassandra/data/rts/40301_feedProducts/rts-40301_feedProducts-ib-1-Data.db (No such file or directory) at org.apache.cassandra.io.compress.CompressedRandomAccessReader.open(CompressedRandomAccessReader.java:46) at org.apache.cassandra.io.util.CompressedSegmentedFile.createReader(CompressedSegmentedFile.java:57) at org.apache.cassandra.io.util.PoolingSegmentedFile.getSegment(PoolingSegmentedFile.java:41)
Large number of pending gossip stage tasks in nodetool tpstats
I'm running cassandra-1.2.8 in a cluster with 45 nodes across three racks. All nodes are well behaved except one. Whenever I start this node, it starts churning CPU. Running nodetool tpstats, I notice that the number of pending gossip stage tasks is constantly increasing [1]. When looking at nodetool gossipinfo, I notice that this node has updated to the latest schema hash, but that it thinks other nodes in the cluster are on the older version. I've tried to drain, decommission, wipe node data, bootstrap, and repair the node. However, the node just started doing the same thing again. Has anyone run into this issue before? Can anyone provide any insight into why this node is the only one in the cluster having problems? Are there any easy fixes? Thank you, Faraaz [1] $ /cassandra/bin/nodetool tpstats Pool NameActive Pending Completed Blocked All time blocked ReadStage 0 0 8 0 0 RequestResponseStage 0 0 49198 0 0 MutationStage 0 0 224286 0 0 ReadRepairStage 0 0 0 0 0 ReplicateOnWriteStage 0 0 0 0 0 GossipStage 1 2213 18 0 0 AntiEntropyStage 0 0 0 0 0 MigrationStage0 0 72 0 0 MemtablePostFlusher 0 0102 0 0 FlushWriter 0 0 99 0 0 MiscStage 0 0 0 0 0 commitlog_archiver0 0 0 0 0 InternalResponseStage 0 0 19 0 0 HintedHandoff 0 0 2 0 0 Message type Dropped RANGE_SLICE 0 READ_REPAIR 0 BINARY 0 READ 0 MUTATION 0 _TRACE 0 REQUEST_RESPONSE 0
Re: Which of these VPS configurations would perform better for Cassandra ?
From what I understood tons of people are running things on ec2, but it could be the instance size is pretty large that it compares to a dedicated server (especially if you go with SSD, it is like 1K/month!) On Tue, Aug 6, 2013 at 3:54 AM, Aaron Morton aa...@thelastpickle.comwrote: how many nodes to start with(2 ok?) ? I'd recommend 3, that will give you some redundancy see http://thelastpickle.com/2011/06/13/Down-For-Me/ Cheers - Aaron Morton Cassandra Consultant New Zealand @aaronmorton http://www.thelastpickle.com On 5/08/2013, at 1:41 AM, Rajkumar Gupta rajkumar@gmail.com wrote: okay, so what should a workable VPS configuration to start with minimum how many nodes to start with(2 ok?) ? Seriously I cannot afford the tensions of colocation setup. My hosting provider provides SSD drives with KVM virtualization.
Re: Which of these VPS configurations would perform better for Cassandra ?
3 node EC2 m1.xlarge is ~ $1000/k month + any incidental costs ( s3 backups, transfer out of the AZ ), etc ) or ~$300/month after a ~$1400 upfront 1 year reservation fee. There are some uncomfortable spots when compaction kicks on concurrently for several large CF's but otherwise its been performant and so far stable using ephemeral raid0 ala ( Datastax's 2.4 AMI ). On Tue, Aug 6, 2013 at 2:15 PM, S Ahmed sahmed1...@gmail.com wrote: From what I understood tons of people are running things on ec2, but it could be the instance size is pretty large that it compares to a dedicated server (especially if you go with SSD, it is like 1K/month!) On Tue, Aug 6, 2013 at 3:54 AM, Aaron Morton aa...@thelastpickle.comwrote: how many nodes to start with(2 ok?) ? I'd recommend 3, that will give you some redundancy see http://thelastpickle.com/2011/06/13/Down-For-Me/ Cheers - Aaron Morton Cassandra Consultant New Zealand @aaronmorton http://www.thelastpickle.com On 5/08/2013, at 1:41 AM, Rajkumar Gupta rajkumar@gmail.com wrote: okay, so what should a workable VPS configuration to start with minimum how many nodes to start with(2 ok?) ? Seriously I cannot afford the tensions of colocation setup. My hosting provider provides SSD drives with KVM virtualization.
Re: Which of these VPS configurations would perform better for Cassandra ?
Amazon seems to much overprice its services. If you look out for a similar size deployment elsewhere like linode or digital ocean(very competitive pricing), you'll notice huge differences. Ok, some services features are extra but may we all don't need them necessarily when you can host on non-dedicated virtual servers on Amazon you can also do it with similar configuration nodes elsewhere too. IMO these huge costs associated with cassandra deployment are too heavy for small startups just starting out. I believe, If you consider a deployment for similar application using MySQL it should be quite cheaper/ affordable(though i'm not exactly sure). Atleast you don't usually create a cluster from the beginning. Probably we made a wrong decision to choose cassandra considering only its technological advantages.