swift 1.9.0 handoff query

Asked by Netmagic Solutions on 2013-07-29

Hi,

We have configured a swift 1.9.0 cluster with 5 zones and 3 replicas, and is functioning without any issues.

I am aware that there has been an improvement in handoff node selection since swift 1.8.0. I see swift-get-nodes output as below :

Server:Port Device 172.26.14.1:6010 HDD02
Server:Port Device 172.26.13.1:6010 HDD03
Server:Port Device 172.26.11.1:6010 HDD05
Server:Port Device 172.26.15.1:6010 HDD02 [Handoff]
Server:Port Device 172.26.12.1:6010 HDD04 [Handoff]
Server:Port Device 172.26.14.1:6010 HDD09 [Handoff]

1) Why is it showing 3 handoffs instead of two as we only have 5 zones.
2) zone 1 is having a copy of the data and another disk is shown from the same zone as handoff

Is this output expected? Is this due to the new handoff selection mechanism or anything is wrong here?

Regards,

Viral

Question information

Language:
English Edit question
Status:
Solved
For:
OpenStack Object Storage (swift) Edit question
Assignee:
No assignee Edit question
Solved by:
Samuel Merritt
Solved:
2013-07-30
Last query:
2013-07-30
Last reply:
2013-07-29
Best Samuel Merritt (torgomatic) said : #1

It's working as expected.

Thanks Samuel Merritt, that solved my question.

serious (guanwen) said : #3

I have three nodes in three zones. Each node has 8 hard disks known to swift as /srv/node/sdc1, sdd1, ...sdj1
I set three replicas.

When I run the command "swift stat", it shows:

Account: AUTH_1c12ca6d87d44d67aa5d63c807babc6a
Containers: 2
   Objects: 1449
     Bytes: 584777968428
Accept-Ranges: bytes
X-Timestamp: 1381982350.81689
Content-Type: text/plain; charset=utf-8

Because I have only three nodes and three replica, I logged in to one node, and "cd /srv/node", and
do the following:

[root@dsb1 node]# find . -type f -name "*data" | wc -l
1449

which is the same as the results from "swift stat". However, there are many objects stored in a hard disk which
is shown as "[handoff]" according to swift-get-nodes /etc/swift/object.ring.gz MyAccount". These objects in the handoff hard disk are counted to the number 1449. If the cluster is operating normally, the objects in handoff node should be cleaned, am I right?

Samuel Merritt (torgomatic) said : #4

Please post new questions as new questions, not comments on existing questions. This question is marked "Solved", which means it won't show up in the list of unanswered questions, so your chances of getting an answer are very low.