tenemos el clúster HDP versión 2.6.5
Cuando miramos name-nodelos registros podemos ver la siguiente advertencia.
2023-02-20 15:56:37,731 INFO namenode.FileJournalManager (FileJournalManager.java:finalizeLogSegment(142)) - Finalizing edits file /hadoop/hdfs/journal/hdfsha/current/edits_inprogress_0000000193594484455 -> /hadoop/hdfs/journal/hdfsha/current/edits_0000000193594484455-0000000193594600017
2023-02-20 15:58:31,377 WARN server.Journal (Journal.java:journal(398)) - Sync of transaction range 193594757835-193594757835 took 1498ms
2023-02-20 15:58:40,617 INFO namenode.FileJournalManager (FileJournalManager.java:finalizeLogSegment(142)) - Finalizing edits file /hadoop/hdfs/journal/hdfsha/current/edits_inprogress_0000000193594600018 -> /hadoop/hdfs/journal/hdfsha/current/edits_0000000193594600018-0000000193594769398
2023-02-20 16:00:39,037 WARN server.Journal (Journal.java:journal(398)) - Sync of transaction range 193594895192-193594895192 took 1371ms
2023-02-20 16:00:42,839 INFO namenode.FileJournalManager (FileJournalManager.java:finalizeLogSegment(142)) - Finalizing edits file /hadoop/hdfs/journal/hdfsha/current/edits_inprogress_0000000193594769399 -> /hadoop/hdfs/journal/hdfsha/current/edits_0000000193594769399-0000000193594899457
2023-02-20 16:01:43,962 WARN server.Journal (Journal.java:journal(398)) - Sync of transaction range 193594954980-193594954980 took 1329ms
2023-02-20 16:02:44,799 INFO namenode.FileJournalManager (FileJournalManager.java:finalizeLogSegment(142)) - Finalizing edits file /hadoop/hdfs/journal/hdfsha/current/edits_inprogress_0000000193594899458 -> /hadoop/hdfs/journal/hdfsha/current/edits_0000000193594899458-0000000193595017147
2023-02-20 16:02:47,129 WARN server.Journal (Journal.java:journal(398)) - Sync of transaction range 193595018764-193595018764 took 1321ms
2023-02-20 16:03:52,763 WARN server.Journal (Journal.java:journal(398)) - Sync of transaction range 193595106645-193595106646 took 1344ms
2023-02-20 16:04:46,965 INFO namenode.FileJournalManager (FileJournalManager.java:finalizeLogSegment(142)) - Finalizing edits file /hadoop/hdfs/journal/hdfsha/current/edits_inprogress_0000000193595017148 -> /hadoop/hdfs/journal/hdfsha/current/edits_0000000193595017148-0000000193595169050
2023-02-20 16:04:56,276 WARN server.Journal (Journal.java:journal(398)) - Sync of transaction range 193595175233-193595175233 took 1678ms
2023-02-20 16:06:01,067 WARN server.Journal (Journal.java:journal(398)) - Sync of transaction range 193595252052-193595252052 took 1265ms
2023-02-20 16:07:06,447 WARN server.Journal (Journal.java:journal(398)) - Sync of transaction range 193595320796-193595320796 took 1273ms
En nuestro clúster HDP, el servicio HDFS incluye 2 name-nodeservicios y 3 journal-Nodes
clústeres incluyen 736 data nodesmáquinas, y el servicio HDFS es el administrador de todosdata-node
Queremos entender ¿cuál es el motivo de la siguiente advertencia?
server.Journal (Journal.java:journal(398)) - Sync of transaction range 193595018764-193595018764 took 1321ms
y cómo evitar estos mensajes mediante una solución proactiva
De lo que encontramos hasta ahora es la siguiente solución:
http://www.hadoopadmin.co.in/hdfs/standby-namenode-is-faling-and-only-one-is-running/
RESOLUTION:
Increase the values of following JournalNode timeout properties:
dfs.qjournal.select-input-streams.timeout.ms = 60000
dfs.qjournal.start-segment.timeout.ms = 60000
dfs.qjournal.write-txns.timeout.ms = 60000