cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Aaron Morton <>
Subject Re: in CompactionExecutor
Date Mon, 21 Feb 2011 12:35:22 GMT
The work around is to disable saving caches, as they may not be correctly saved. However you
do not have access to change the settings for the system CFs. I imagine it may only be an
issue if a node is rebooted as it may load stale caches.

I think it's windows only as the function used to rename the saved cache says it has different
operation on different OS's and
there seems to be a number of people complaining about the behavior on windows. 

On my ubuntu desktop I ran a test where a user CF safes it's caches every 5 seconds and all
was well. 

32bit Win XP (sp3) may not fall into the list of supported OS's for cassandra, but we may
confirm how it works on a more modern Windows version.
Let me when you have created the bug. 



On 22 Feb, 2011,at 01:14 AM, ruslan usifov <> wrote:

I have a question. How do you think this happens only on windows(on this platform i don't
worry, because this is only test platform) or everywhere (linux)? And how dangerous this error,
or at first time i may simply ignore it?

2011/2/21 Aaron Morton <>


On 22 Feb, 2011,at 12:55 AM, ruslan usifov <> wrote:

2011/2/21 Aaron Morton <>
The code creates a new .tmp file in the saved_caches directory and then renames it to a non
.tmp file name, so there is nothing else with a handle open. The rename is to an existing
file though. 

Ruslan can you please raise a bug against 0.7.2 for this and include the platform. 

You mean in JIRA?
This happens on WindowsXP (SP3) 32 bit

  • Unnamed multipart/alternative (inline, None, 0 bytes)
    • Unnamed multipart/related (inline, None, 0 bytes)
View raw message