config servers traffic

[tetlika]

hi!

i have 3 config servers in my deployment, and noticed that if the
balancing is on - the outgoing traffic of the first one config (that
which goes first in config of mongoses) is 10Mbits while on other
config servers it's around few kilobits,

if to turn off balancing - the traffic becomes the same on all
config servers (few kilobits)

the sharding seems work ok, i'm monitoring such situation from the
very beggining of sharding (more than a year), and looks it's not
documented

is it normal?

[Andre de Frere]

Hi,

This is expected behaviour.  Because the order of the config servers in the --configdb parameter of the mongos matters, and the mongos will always use the first of those config servers when addressing the configDB, there will be more traffic to the first config server in the string.  With the balancing switched off, no moveChunk operations will take place and the chance of getting a staleConfigDetected (and therefore needing to reload the config from the configDB) reduced to nearly nothing.  This means that the traffic to the first config server will be much reduced.

Please note, you cannot simply randomise the list of config servers per mongos.  The --configdb string needs to be the same across all mongos'.

Regards,

André

[Mykola Shenhera]

Hi,

We have the same issue with very high traffic from first config server to mongos: 750GB pre day. While traffic from the rest two config servers only 110MB. Can someone explain the cause of it and what can be done to reduce the traffic?

--
Thanks you,
Mykola

[tetlika]

Mykola, looks this is normal - see previous reply from Andre de Frere in this thread

Вівторок, 15 липня 2014 р. 16:01:55 UTC+3 користувач Mykola Shenhera написав:

[Mykola Shenhera]

It's OK until deploy mongo on AWS and choosing instance type for config server. According to MongoDB docs t1.micro is sufficient. However taking to consideration low network performance of t1.micro and high traffic of mongo server, instance with better performance should be used. And whole setup of mongo sharded cluster on AWS become too expensive.

And taking to consideration that total traffic on shard node is 3G it's strange to see 750 GB on config server.

[tetlika]

agree, and we use m3.large for configs because of that

Вівторок, 15 липня 2014 р. 19:54:40 UTC+3 користувач Mykola Shenhera написав:

[tetlika]

P.S

as high traffic is noticed just during when balancing is on, maybe you could try scheduled balancing (when the load is the lowest)

that will possibly save some $$$ in that case

Вівторок, 15 липня 2014 р. 19:54:40 UTC+3 користувач Mykola Shenhera написав:

[Asya Kamsky]

Where does it say that t1.micro is sufficient for config servers?  That sounds incorrect/out of date and should be fixed.

--
You received this message because you are subscribed to the Google Groups "mongodb-user"
group.
 
For other MongoDB technical support options, see: http://www.mongodb.org/about/support/.
---
You received this message because you are subscribed to the Google Groups "mongodb-user" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mongodb-user...@googlegroups.com.
To post to this group, send email to mongod...@googlegroups.com.
Visit this group at http://groups.google.com/group/mongodb-user.

To view this discussion on the web visit https://groups.google.com/d/msgid/mongodb-user/f6fd81c5-c629-42cc-80a6-2a0ecf3dd5f1%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.