Hydra Cache

Unaccessible node removal does not seem to work correctly

Details

  • Type: Bug Bug
  • Status: Resolved Resolved
  • Priority: Major Major
  • Resolution: Fixed
  • Affects Version/s: 1.0RC1
  • Fix Version/s: 1.0 GA
  • Component/s: Client
  • Labels:
    None
  • Number of attachments :
    0

Description

Unaccessible node removal does not seem to work correctly. The debug msg is displayed stating the node has been removed from the ring after connection refusal, however the subsequent admin related operation seem still hitting the unreachable node.

Activity

Ascending order - Click to sort in descending order
Nick Zhu made changes -
Field Original Value New Value
Assignee Nick Zhu [ aucifer_nick ]
Nick Zhu made changes -
Original Estimate 0 minutes [ 0 ]
Remaining Estimate 0 minutes [ 0 ]
Nick Zhu made changes -
Status Open [ 1 ] In Progress [ 3 ]
Hide
Permalink
Nick Zhu added a comment -

The sendMessage method for binary data payload knows to remove the node from the ring however all JSON related admin methods do not execute the same logic to remove unaccessible node thus in hconsole when data and admin clients were separately created, the admin client keeps hitting the off line node even when it has previously failed connecting to the node

Show
Nick Zhu added a comment - The sendMessage method for binary data payload knows to remove the node from the ring however all JSON related admin methods do not execute the same logic to remove unaccessible node thus in hconsole when data and admin clients were separately created, the admin client keeps hitting the off line node even when it has previously failed connecting to the node
Hide
Permalink
Nick Zhu added a comment -

After further investigation I found the problem is actually in the algorithm used to pick a random node for admin requests since the initial seed server list was never updated for the client it ends up picking the offline node. The client should update the seed server list after getting the full picture from the seed server initially.

Show
Nick Zhu added a comment - After further investigation I found the problem is actually in the algorithm used to pick a random node for admin requests since the initial seed server list was never updated for the client it ends up picking the offline node. The client should update the seed server list after getting the full picture from the seed server initially.
Nick Zhu made changes -
Status In Progress [ 3 ] Resolved [ 5 ]
Resolution Fixed [ 1 ]

People

  • Assignee:
    Nick Zhu
    Reporter:
    Nick Zhu
Vote (0)
Watch (0)

Dates

  • Created:
    Updated:
    Resolved: