> This would require that Cassandra run as root on Linux systems, as 'man
> mlockall' states:
IIRC, mlock() (as opposed to mlockall()) does not require root
privileges - but is subject to resource limitations.
However, given a lack of control of how memory is allocated in the JVM
I suppose mlock
Okay, so things were pretty messed up. I shut down all the new nodes,
then the old nodes started doing the half the ring is down garbage which
pretty much requires a full restart of everything. So I had to shut
everything down, then bring the seed back, then the rest of the nodes,
so they finally
yes, you need to maintain the secondary index your self. Send a
batch_mutation and write the article and website article colums at the
same time. I think your safe up to a large number of cols, say
1M Not sure, may try to track the info down one day.AOn 16 Jul, 2010,at 03:39 PM, S Ahmed wrote:S
On 2010-07-16 01:57, Dave Viner wrote:
> I am no expert... but parts seem accurate, parts not.
>
> "Cassandra stores four or five dimension associated arrays"
> not sure what you're counting as a dimension of the associated array,
> but here are the 2 associative array-like syntaxes:
>
> ColumnFa
The column names are arbitrary strings, so it's not obvious what the
"next" value should be at any step. So, I just set the start of the
next page to the end of the last page and eliminate the duplicate
value when joining the 2 pages together.
The paging direction does not matter in my case, as I
So am I to keep track on the # of columns for a given key in CF
WebsiteArticle? i.e. if I want to do a get_slice for the first 10 OR last
10 (I would need to know the count to get the last 10).
>>Am assuming RP. There are some recommendations on the number of cols per
key, in the millions I think
You should make sure that your directions and interval endpoints are chosen
correctly. I recall the semantics of the call being like an old-school for
with the descending flag as a step of +1 or -1.
--
Spelling by mobile.
On Jul 15, 2010, at 20:19, Ilya Maykov wrote:
> Hi all,
>
> I'm tryi
Hi all,
I'm trying to debug some pretty weird behavior when paginating through
a ColumnFamily with get_slice(). It basically looks like Cassandra
does not respect the limit parameter in the SlicePredicate, sometimes
returning more than limit columns. It also sometimes silently drops
columns. I'm r
I am no expert... but parts seem accurate, parts not.
"Cassandra stores four or five dimension associated arrays"
not sure what you're counting as a dimension of the associated array, but
here are the 2 associative array-like syntaxes:
ColumnFamily[row-key][column-name] = value1
ColumnFamily[row-
i saw this in the kernel log: jsvc uses 32-bit capabilitie. Is this right?
our server is
Linux 2.6.32-23-generic #37-Ubuntu SMP Fri Jun 11 08:03:28 UTC 2010 x86_64
GNU/Linux
On Jul 15, 2010, at 11:04 AM, Claire Chang wrote:
> I am using Random Partitioner. The other 2 nodes are working fine.
You could build a secondary index, e.g.CFArticles : {article_id1 : {}article_id2 : {}}CFWebsiteArticle : {website_id1 : { time_uuid : article_id1, time_uuid2 : article_id2}}when you want to get the last 10 for a website, get_slice from the WebsiteArticle CF then multi get from Articles. Am assuming
Hi,
I am writing a scalability chapter in a book and I need to mention
Apache Cassandra although it's just a mention. Still I would not like
to be sloppy and would like to get verification whether my summary is
accurate. "Cassandra stores four or five dimension associated arrays.
The first dimensi
On Thu, Jul 15, 2010 at 5:46 PM, Clint Byrum wrote:
> One other approach that works on Linux is to use HugeTLB. This post details
> the process for doing so with a jvm:
>
> http://andrigoss.blogspot.com/2008/02/jvm-performance-tuning.html
>
> Basically when mmapping using HUGETLB you don't have t
On Jul 15, 2010, at 2:52 PM, Jonathan Ellis wrote:
> On Thu, Jul 15, 2010 at 3:56 PM, Carlos Alvarez wrote:
>> On Thu, Jul 15, 2010 at 2:01 PM, Jonathan Ellis wrote:
>>> The main problem is not the syscall so much as Java insisting on
>>> zeroing out any buffer you create, which is a big hit to
On Thu, Jul 15, 2010 at 5:29 PM, Mubarak Seyed wrote:
> Just want to verify with group that what i am doing wrt RF is correct.
> 1. Nodes were running with RF=2
> 2. Stopped all the nodes, changed the RF to 4
> 3. Started all the nodes, verify the cluster ring using nodetool, all the
> nodes are
Just want to verify with group that what i am doing wrt RF is correct.
1. Nodes were running with RF=2
2. Stopped all the nodes, changed the RF to 4
3. Started all the nodes, verify the cluster ring using nodetool, all the
nodes are part of cluster
4. Ran nodetool repair on all the nodes
5. Ran n
On Thu, Jul 15, 2010 at 3:56 PM, Carlos Alvarez wrote:
> On Thu, Jul 15, 2010 at 2:01 PM, Jonathan Ellis wrote:
>> The main problem is not the syscall so much as Java insisting on
>> zeroing out any buffer you create, which is a big hit to performance
>> when you're allocating buffers for file i/
Given a CF like:
Articles : {
key1 : { title:"some title", body: "this is my article body...", },
key1 : { title:"some title", body: "this is my article body...", }
}
Now these articles could be for different websites e.g. www.website1.com,
www.website2.com
If I want to get the
Benjamin,
Ah, thanks for clarifying that.
key sorting is changing in .7 I believe to support a binary array?
On Thu, Jul 15, 2010 at 3:26 PM, Benjamin Black wrote:
> Keys are always sorted (in 0.6) as UTF8 strings. The CompareWith
> applies to _columns_ within rows, _not_ to row keys.
>
> On
On Thu, Jul 15, 2010 at 2:01 PM, Jonathan Ellis wrote:
> The main problem is not the syscall so much as Java insisting on
> zeroing out any buffer you create, which is a big hit to performance
> when you're allocating buffers for file i/o on each request instead of
> just mmaping things. Re-using
On Thu, Jul 15, 2010 at 3:28 PM, Anthony Molinaro
wrote:
> Is the fact that 2 new nodes are in the range messing it up?
Probably.
> And if so
> how do I recover (I'm thinking, shutdown new nodes 2,3,4,5, the bringing
> up nodes 2,4, waiting for them to finish, then bringing up 3,5?).
Yes.
You
Oh, and looking at the load on the new machines it appears that
New 2 and New 6 have gotten some data (although neither is in the ring
yet). Not sure if that clears anything up though.
-Anthony
On Thu, Jul 15, 2010 at 01:28:06PM -0700, Anthony Molinaro wrote:
> This is a cluster which is horri
This is a cluster which is horribly imbalanced because I didn't assign
initial tokens, so I'm adding 6 nodes with tokens according to the operations
page (ie, i * (2^127/N) with N = 6).
So here's what the ring will look like when bootstrap finishes
151901684708361811491018697
Keys are always sorted (in 0.6) as UTF8 strings. The CompareWith
applies to _columns_ within rows, _not_ to row keys.
On Wed, Jul 14, 2010 at 1:44 PM, S Ahmed wrote:
> Where is the link that describes the various key types and their impact on
> sorting? (I believe I read it before, can't seem to
On Thu, Jul 15, 2010 at 1:54 PM, B. Todd Burruss wrote:
> if i have N=3 and run nodetool repair on node X. i assume that merkle
> trees (at a minimum) are calculated on nodes X, X+1, and X+2 (since
> N=3). when the repair is finished are nodes X, X+1, and X+2 all in sync
> with respect to node X
> I'm convinced. :) See comments on
> https://issues.apache.org/jira/browse/CASSANDRA-1214
Noted :) To be clear I only mentioned it as an acknowledgement that
everyone didn't necessarily agree with what I was saying.
> The main problem is not the syscall so much as Java insisting on
> zeroing ou
> If you could can you please share the command line function (to load TSV)?
There is no command line function ... you have to write code for this.
> and Can you please help me on storing storage-conf.xml on HDFS part?
As I said. Maybe you better start with a simpler scenario and leave
out HDFS
if i have N=3 and run nodetool repair on node X. i assume that merkle
trees (at a minimum) are calculated on nodes X, X+1, and X+2 (since
N=3). when the repair is finished are nodes X, X+1, and X+2 all in sync
with respect to node X's data? or does X have the latest data and X+1
and X+2 still in
I am using Random Partitioner. The other 2 nodes are working fine. There are no
Errors in the log files for the 2 good nodes.
There were no log messages within 30 minutes before the exception occurs. Here
is the last log statement before the exception occurred.
INFO [COMPACTION-POOL:1] 2010-07
I am trying to set up a 3 node cluster. RF=3 and CL=1 for most of the request.
The initial seeding took about 1 hour to complete which loaded each node with
2G of data. After the seeding completed, one node started having this exception
and hung. Read/Write with CL=ALL timed out but CL=QUORUM wa
Hi Torsten,
If you could can you please share the command line function (to load TSV)?
and Can you please help me on storing storage-conf.xml on HDFS part?
Thanks,
Mubarak
On Tue, Jul 13, 2010 at 1:27 AM, Torsten Curdt wrote:
> On Tue, Jul 13, 2010 at 04:35, Mubarak Seyed
> wrote:
> > Where c
On Thu, Jul 15, 2010 at 11:41 AM, Peter Schuller
wrote:
> Not really. That is, the intent of mmap is to let the OS dynamically
> choose what gets swapped in and out. The practical problem is that the
> OS will often tend to swap too much. I got the impression jbellis
> wasn't convinced, but my ane
Yes, I think current HintedHandOff implementation in 0.6.x cannot support
large hints, it is a risk in a production system.
On Tue, Jun 29, 2010 at 12:31 AM, albert_e wrote:
> In 0.6.2, HH sending MUTATION message using the same OutboundTcpConnection
> with READ message. When HH transfering big
I found, for large dataset, long-term random reading test, the performance
with mmap is very bad.
See the attached chart in
https://issues.apache.org/jira/browse/CASSANDRA-1214.
On Fri, Jul 16, 2010 at 12:41 AM, Peter Schuller <
peter.schul...@infidyne.com> wrote:
> > Can someone please explain t
> Can someone please explain the mmap issue.
> mmap is default for all storage files for 64bit machines.
> according to this case https://issues.apache.org/jira/browse/CASSANDRA-1214
> it might not be a good thing.
> Is it right to say that you should use mmap only if your MAX expected data
> is sm
Do you think a composite key using a key type of Bytes would work?
How many bytes can it be?
public static byte [] createRowKey(int websiteid, long stamp)
throws Exception {
byte [] websiteidBytes = Bytes.toBytes(websiteid);
byte [] stampBytes = Bytes.toBytes(stamp);
return Bytes.add(websi
Hi,
I am using OrderPreservingPartitioner, and my keys are integers which are
stored as strings, I want to manually assign token values equal to my key
values such that data is equally distributed.
So for this to work, I want to convert the token and key strings to integers
before doing compareTo
Well I'm not talking about a specific column family here, as ALL my column
families will have content that is specific to a certain website, so I need
a strategy that I will use on almost all my column families.
On Wed, Jul 14, 2010 at 9:20 PM, Schubert Zhang wrote:
> for your apps, how about th
Did you add a new node to the cluster at the time you restarted it?
If not, I would think that each node already had a token that would
make such a collision impossible, unless we have a new bug to
troubleshoot.
Gary.
On Wed, Jul 14, 2010 at 20:46, Mubarak Seyed wrote:
> The cluster nodes were r
Short answer: yes, this is normal.
Longer answer: this was discussed at length on this list a few days
ago, check the archives.
On Wed, Jul 14, 2010 at 10:55 PM, Hendro Kaskus
wrote:
> Hi everyone,
>
> I'm newbie to Cassandra :D.. I try to insert data from MySQL to Cassandra.
> Data dump from My
Can someone please explain the mmap issue.
mmap is default for all storage files for 64bit machines.
according to this case
https://issues.apache.org/jira/browse/CASSANDRA-1214it might not be a
good thing.
Is it right to say that you should use mmap only if your MAX expected data
is smaller then th
It could be that your Cassandra nodes haven't full compacted yet.
On Thu, Jul 15, 2010 at 5:55 AM, Hendro Kaskus wrote:
> Hi everyone,
>
> I'm newbie to Cassandra :D.. I try to insert data from MySQL to Cassandra.
> Data dump from MySQL is about 11 MB (64716 records). But when i'm insert to
> Cas
42 matches
Mail list logo
