Hi,

 

Could you share the full logs with the exception?

 

There is an interesting  JIRA ticket that could be the reason for this [1]

 

[1] - https://issues.apache.org/jira/browse/IGNITE-13373

 

 

 

From: shivakumar
Sent: Thursday, December 3, 2020 10:51 AM
To: [email protected]
Subject: Could not clear historyMap due to WAL reservation on cp

 

Hi

I have deployed 5 node Ignite cluster on K8S with persistence enabled

(version 2.9.0 on java 11)

I started ingesting data to 3 tables and after ingesting large amount of

data using JDBC batch insertion (around 20 million records to each of 3

tables with backup set to 1), now i connected to visor shell (from one of

the pod which i deployed just to use as visor shell) using the same ignite

config file which is used for ignite servers and after visor shell connects

to ignite cluster the unwanted wal record cleanup stopped (which should run

post checkpoint ) and WAL started growing linearly as there is continuous

data ingestion. This is making WAL disk run out of space and pods crashes.

 

when i see logs there are continuous warning messages saying

Could not clear historyMap due to WAL reservation on cp: CheckpointEntry

[id=e8bb9c22-0709-416f-88d6-16c5ca534024, timestamp=1606979158669,

ptr=FileWALPointer [idx=1468, fileOff=45255321, len=9857]], history map size

is 4

 

 

and if i see checkpoint finish messages

 

Checkpoint finished [cpId=ca254956-5550-45d6-87c5-892b7e07b13b,

pages=494933, markPos=FileWALPointer [idx=1472, fileOff=72673736, len=9857],

walSegmentsCleared=0, walSegmentsCovered=[1470 - 1471], markDuration=464ms,

pagesWrite=8558ms, fsync=3597ms, total=14027ms]

 

here you can see walSegmentsCleared=0   means there are no WAl segments

cleared even after checkpoint, not sure what is causing this behaviour.

 

we are ingesting very large data (~25mb/s)

please someone help in this issue

 

 

regards,

Shiva

 

 

 

 

 

--

Sent from: http://apache-ignite-users.70518.x6.nabble.com/

 

Reply via email to