p*2
2 楼
mongo网站上说mongo很适合做logging,有大牛谈谈吗?我怎么感觉很不适合呢。
c*e
5 楼
"mongo要用内存,数据量大了放不下"
Log is append-only, not a big issue here regarding memory. It’s true
replica set in Mongo is quite expensive but you cannot really get away if
you want availability.
Efficient logging requires hierarchical design. You can aggregate warning/
errors/metrics to centralized storage while leaving many verbose ones on
node. It's not necessary to put everything on persistent storage and it's
expensive anyway. As long as you can search across nodes or even have global
tracing, debug/diagnostics should be easier.
Log is append-only, not a big issue here regarding memory. It’s true
replica set in Mongo is quite expensive but you cannot really get away if
you want availability.
Efficient logging requires hierarchical design. You can aggregate warning/
errors/metrics to centralized storage while leaving many verbose ones on
node. It's not necessary to put everything on persistent storage and it's
expensive anyway. As long as you can search across nodes or even have global
tracing, debug/diagnostics should be easier.
p*2
6 楼
global
log 是 append的,但是query 的时候就有问题了吧?mongo aggreate只能支持16M,数
据大了根本没法用,mapreduce也很慢。感觉数据量大了,performance总是有问题。如
果不做dynamic query的话,那不如放其他地方了。
另外跟cassandra, hbase比有什么优势呢?从写log来说。
【在 c****e 的大作中提到】
: "mongo要用内存,数据量大了放不下"
: Log is append-only, not a big issue here regarding memory. It’s true
: replica set in Mongo is quite expensive but you cannot really get away if
: you want availability.
: Efficient logging requires hierarchical design. You can aggregate warning/
: errors/metrics to centralized storage while leaving many verbose ones on
: node. It's not necessary to put everything on persistent storage and it's
: expensive anyway. As long as you can search across nodes or even have global
: tracing, debug/diagnostics should be easier.
c*e
7 楼
没有太多优势。
Don't use mongodb's JS mapreduce, slow.
https://engineering.groupon.com/2013/big-data/mongodb-mapreduce-with-hadoop/
问题是你在折腾什么啊? Logstash, fluentd现成的为什么不用?
【在 p*****2 的大作中提到】
:
: global
: log 是 append的,但是query 的时候就有问题了吧?mongo aggreate只能支持16M,数
: 据大了根本没法用,mapreduce也很慢。感觉数据量大了,performance总是有问题。如
: 果不做dynamic query的话,那不如放其他地方了。
: 另外跟cassandra, hbase比有什么优势呢?从写log来说。
Don't use mongodb's JS mapreduce, slow.
https://engineering.groupon.com/2013/big-data/mongodb-mapreduce-with-hadoop/
问题是你在折腾什么啊? Logstash, fluentd现成的为什么不用?
【在 p*****2 的大作中提到】
:
: global
: log 是 append的,但是query 的时候就有问题了吧?mongo aggreate只能支持16M,数
: 据大了根本没法用,mapreduce也很慢。感觉数据量大了,performance总是有问题。如
: 果不做dynamic query的话,那不如放其他地方了。
: 另外跟cassandra, hbase比有什么优势呢?从写log来说。
p*2
8 楼
hadoop/
大牛。不是我折腾。是我老板觉得mongo特别适合做logging。我现在是想找理由说服
team放弃mongo用Logstash。
【在 c****e 的大作中提到】
: 没有太多优势。
: Don't use mongodb's JS mapreduce, slow.
: https://engineering.groupon.com/2013/big-data/mongodb-mapreduce-with-hadoop/
: 问题是你在折腾什么啊? Logstash, fluentd现成的为什么不用?
P*i
10 楼
都是Ruby?
hadoop/
【在 c****e 的大作中提到】
: 没有太多优势。
: Don't use mongodb's JS mapreduce, slow.
: https://engineering.groupon.com/2013/big-data/mongodb-mapreduce-with-hadoop/
: 问题是你在折腾什么啊? Logstash, fluentd现成的为什么不用?
hadoop/
【在 c****e 的大作中提到】
: 没有太多优势。
: Don't use mongodb's JS mapreduce, slow.
: https://engineering.groupon.com/2013/big-data/mongodb-mapreduce-with-hadoop/
: 问题是你在折腾什么啊? Logstash, fluentd现成的为什么不用?
z*e
14 楼
mongodb的优势是跟db有很大重合
让人一看就觉得很不靠谱
如果是那样的话,我回头去用db好了
让人一看就觉得很不靠谱
如果是那样的话,我回头去用db好了
相关阅读
Core Java2 Notes (4)How to compile a Java program with import my own package in UNIX除了sun的服务器,哪里还有java的api?J2EE vs .NET (上)(z)List, LinkedList and VectorMwTerm 支持拼音输入, GB/BIG5 简/繁体 Release 1HTMLHelp版的,偶的最爱Re: 除了sun的服务器,哪里还有java的api?.net vs. J2ee(zhuan)Core Java2 Notes (6)Code InstrumenterUNIX build using ANT!!!can applet implements runnable?Brainbench ejb examJava Telnet Client Implementation v0.991BetaCore Java2 Notes (3)Core Java2 Notes (1)first post in 2003[转载] zz 王森:程序设计师真情忏悔录(修正完整版)Want take SCEA警告TeacherWei