You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@ignite.apache.org by sk...@apache.org on 2023/04/13 07:38:05 UTC
[ignite-3] branch main updated: IGNITE-15087 Removed outdated TODO
This is an automated email from the ASF dual-hosted git repository.
sk0x50 pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/ignite-3.git
The following commit(s) were added to refs/heads/main by this push:
new eaeb43db00 IGNITE-15087 Removed outdated TODO
eaeb43db00 is described below
commit eaeb43db00a03a3325c211ab64d83c3ec603341c
Author: Slava Koptilin <sl...@gmail.com>
AuthorDate: Thu Apr 13 10:37:53 2023 +0300
IGNITE-15087 Removed outdated TODO
---
modules/transactions/README.md | 6 ++----
1 file changed, 2 insertions(+), 4 deletions(-)
diff --git a/modules/transactions/README.md b/modules/transactions/README.md
index 58c0ed6a55..558b3c4778 100644
--- a/modules/transactions/README.md
+++ b/modules/transactions/README.md
@@ -190,16 +190,14 @@ The locking rules are same as for get/put operations.
Then values are removed from indexes on step 2, they are written as tombstones to avoid read inconsistency and should be
cleaned up after tx finish.
-TODO IGNITE-15087: tx example flow with enabled index(es)
-
# Failover handling
Failover protocol is similar to Ignite 2 with a main difference: until tx is sure it can commit or rollback, it holds
its locks. This means in the case of split-brain, some keys will be locked until split-brain situation is resolved and
tx recovery protocol will converge. Consult a 2PC paper for details when it's possible, for example <sup id="a3">[3](#f3)</sup>
-## Leaserholder fail
+## Leaseholder fail
If a tx is not started to COMMIT, the coordinator reverts a transaction on remaining leaseholders.
-Then a new leasholder is elected, it checks for its pending transactions and asks a coordinator if it's possible to commit.
+Then a new leaseholder is elected, it checks for its pending transactions and asks a coordinator if it's possible to commit.
## Coordinator fail
Broadcast recovery (various strategies are possible: via gossip or dedicated node) is necessary (because we don't have