When the row is stored on disk as SSTable, the complexity of getting a row
is constant, as it always know where to get the row by in-memory indices.
When the row is stored in memory as memtable, it is stored as skip list[1].
The complexity is O(logN). N is the total number of rows in the skip li
On Fri, Dec 24, 2010 at 6:42 PM, Zhu Han wrote:
> When the row is stored on disk as SSTable, the complexity of getting a row
> is constant, as it always know where to get the row by in-memory indices.
>
BTW: not the whole indices are kept in memory, just part of them are. This
is controlled by "
>> In any case: Monitoring disk-space is very very important.
>
> So, why doesn't cassandra monitor it itself and stop accepting writes if it
> runs out of space?
For one thing, it's non-trivial to do accurately because disk space
usage varies over time due to background compaction and/or
anti-co
>> When the row is stored on disk as SSTable, the complexity of getting a row
>> is constant, as it always know where to get the row by in-memory indices.
>
> BTW: not the whole indices are kept in memory, just part of them are. This
> is controlled by "IndexInterval". That is, 1/IndexInterval of
On Dec 23, 2010, at 12:34, Timo Nentwig wrote:
> On Dec 23, 2010, at 9:34, Timo Nentwig wrote:
>
>> I was about to add a secondary index (which apparently failed) to existing
>> data. When I restarted the node it crashed (!) with:
>
> It crashed because it ran out of heap space (2G). So I incr
On Fri, Dec 24, 2010 at 4:42 AM, Zhu Han wrote:
> When the row is stored on disk as SSTable, the complexity of getting a row
> is constant, as it always know where to get the row by in-memory indices.
Technically, it's O(log N) because of the binary search on the in-memory index.
--
Jonathan El
Yep. I forgot about the binary search part.
Thank you!
regards,
hanzhu
On Fri, Dec 24, 2010 at 9:35 PM, Jonathan Ellis wrote:
> On Fri, Dec 24, 2010 at 4:42 AM, Zhu Han wrote:
> > When the row is stored on disk as SSTable, the complexity of getting a
> row
> > is constant, as it always know
Sorry but I am not sure how to answer all the question that you have posed
since a lot of the stuff I am working with is quite new to me and I haven't
use many of the tools that are talked about but I will try my best to answer
the question to the best of my knowledge. I am trying to get the cassan
I am using the Hadoop interface with Cassandra. Is it possible to line up
partitions or splits of two different column families to be on the same node? I
am doing this for data locality reasons. I want to read all the data from a
split of column family A and a split from column family B into mem
Hum, very strange.
More what I was trying to get at was: did the process truly die or was it
just non-responsive and looking like it was dead? It would be very strange
if the actual process was dying without any warnings in the logs. Presumably
you are running bin/cassandra *without* the -f option
I am running the bin/cassandra with the -f option and it does seem to fully
die and not just stalling.
I have also tried using the cassandra-cli to create keyspace and it works
for a little bit and then will die slightly after accepting the request the
vmstat after it dies is as follows:
procs -
yes, it looks like the workaround of using an initial token of 1 works.
thanks,
-mike
On Dec 23, 2010, at 3:47 PM, Jonathan Ellis wrote:
> On Thu, Dec 23, 2010 at 3:00 PM, mike dooley wrote:
>> DEBUG [pool-1-thread-4] 2010-12-23 12:54:26,958 StorageProxy.java (line 597)
>> restricted ranges fo
One last clarification given you are running with -f, fully die=return to
command prompt with no action on your part? If you ctrl-c from Cassandra
when running in foreground mode (ie with f), the process WILL be killed.
Try running in background mode (without the f).
Removing the contents o
Ho ho holy crap, even on x-mas eve. That's right, it's 0.7.0 RC #3, and
now everyone has a little something under their tree.
Now, if anyone asks, I took the time again to remind you that this isn't
an official stable release, that there might be bugs, that you should
proceed with caution, yada,
> You should still exercise caution
> upgrading anything that matters, but now is the time to test. Please.
For those interested in a distributed test harness, several in the Bay Area
Cassandra community have started one:
https://issues.apache.org/jira/browse/CASSANDRA-1859
On Dec 24, 2010, at
thanks all - and just to clarify the cost of getting a column (and a column
given a SuperColumn) is also O(log N) ?
From: Zhu Han
To: user@cassandra.apache.org
Sent: Sat, 25 December, 2010 3:16:08 AM
Subject: Re: complexity
Yep. I forgot about the binary sear
Hi,
I am new to Cassandra and Hector and have a basic question on fetching
multiple columns of a row that have mixed data types.
So, I am basically doing equivalent of "Select dateCol1, dateCol2,
stringCol1, intCol1 from a_table where key in (?, ?, ?...)"
My question is do I need to do the conve
In the case where you have different value types within the same
slice, yes, you must handle the conversion yourself and
ByteBufferSerializer is the easiest way to do that.
On Fri, Dec 24, 2010 at 10:44 PM, Roshan Dawrani
wrote:
> Hi,
>
> I am new to Cassandra and Hector and have a basic question
RPMs are available via rpm.riptano.com
There have been a few more changes to the package structure since rc2,
primarily to bring the Riptano RPM in-line with the spec file included
in the source tree (see below for details). This should be the last
round of naming and version changes - apologies f
Or you can specify the types with the column name instead of doing a slice.
On Fri, Dec 24, 2010 at 11:23 PM, Nate McCall wrote:
> In the case where you have different value types within the same
> slice, yes, you must handle the conversion yourself and
> ByteBufferSerializer is the easiest way t
Can you please point me an example of that or some Hector API documentation?
I am currently using IndexedSliceRangeQuery to filter on a 2ndary index and
fetching all columns of the matching rows - using range start/end = null. Can
the option you suggest be used with the 2ndry index filtering? I
With a very small amount of memory, the Cassandra process may be getting
killed by the Linux OOM killer, which should result in a log message to the
kernel logs. See
http://stackoverflow.com/questions/624857/finding-which-process-was-killed-by-linux-oom-killerto
locate the error if it exists.
On F
22 matches
Mail list logo
