search

ibm-research / swifthlm

SwiftHLM - a middleware for using OpenStack Swift with tape and other hight latency media storage backends
Apache License 2.0
14 stars 4 forks source link

when one swift node down, object recall is pending #13

Open DaisyHou opened 6 years ago

DaisyHou commented 6 years ago

Description: when one swift node in the cluster is down, object recall operation will be pending Steps:

  1. Put a new object on the container with swift restful API to the cluster
  2. Migrate the object using swifthlm API.
  3. reboot one swift node which having one replica of this object.
  4. recall the object using swifthlm API.
  5. check the request status is pending. "/20180330072055.321/recall/AUTH_8687d6bf9b8b4508a6ad2282eb2c977c/cont_ei/0/obj1--pending",
DaisyHou commented 6 years ago

after recovery the swift node which is rebooted, the recall request is done. But the status is not changed to 'premigrated', it's still 'migrated'. check request: curl -s -H "$Auth" http://$IP/hlm/v1/requests/$Account/cont_ei/obj1 | python -m json.tool [ "There are no pending or failed SwiftHLM requests." ]

check status: curl -s -H "$Auth" http://$IP/hlm/v1/status/$Account/cont_ei/obj1 | python -m json.tool { "/AUTH_8687d6bf9b8b4508a6ad2282eb2c977c/cont_ei/obj1": "migrated" }

check status by ltfsdm on the all the nodes which have the obj1's replicas: state size blocks tape id file name p 258894 520 A00030L5 /mnt/xfs4swift/objects/599/c98/95d25e5d726c78c0a7a127cc3d3adc98/1522393875.07578.data