Java crash on M4 Macs causing OCIS docker container restart loop

[prohtex]

I have a new WOPI deployment running in a Ubuntu 24.04.1 VM on an M4 Mac host. After finishing the deployment, the Tika container keeps restarting with a Java crash.

# docker container list
CONTAINER ID   IMAGE                          COMMAND                  CREATED          STATUS                            PORTS                                                                      NAMES
375646d0829c   owncloud/ocis-rolling:latest   "/bin/sh -c 'ocis co…"   13 minutes ago   Up 3 minutes                      9200/tcp                                                                   ocis_full-collaboration-1
58f6ff00938d   owncloud/ocis-rolling:latest   "/bin/sh -c 'ocis in…"   13 minutes ago   Up 10 seconds                     9200/tcp                                                                   ocis_full-ocis-1
6898ae39998a   traefik:v3.2.2                 "/entrypoint.sh --lo…"   13 minutes ago   Up 13 minutes                     0.0.0.0:80->80/tcp, :::80->80/tcp, 0.0.0.0:443->443/tcp, :::443->443/tcp   ocis_full-traefik-1
6d7fcc2442e1   collabora/code:24.04.10.2.1    "bash -c 'coolconfig…"   13 minutes ago   Up 13 minutes (healthy)           9980/tcp                                                                   ocis_full-collabora-1
3eaa53f68e32   apache/tika:latest-full        "/bin/sh -c 'exec ja…"   13 minutes ago   Restarting (127) 32 seconds ago         

The ocis-rolling container and apache containers last a matter of seconds before restarting.

The logs look like this:

$ sudo docker logs --tail 50 --follow --timestamps 3eaa53f68e32
2025-02-05T04:50:12.047498970Z # /tmp/hs_err_pid1.log
2025-02-05T04:50:12.048156460Z [0.010s][warning][os] Loading hsdis library failed
2025-02-05T04:50:12.050759802Z #
2025-02-05T04:50:12.050772424Z # The crash happened outside the Java Virtual Machine in native code.
2025-02-05T04:50:12.050774132Z # See problematic frame for where to report the bug.
2025-02-05T04:50:12.050787921Z #
2025-02-05T04:50:12.050840285Z 
2025-02-05T04:50:12.050863697Z [error occurred during error reporting (), id 0x5, SIGTRAP (0x5) at pc=0x0000fbc1be941b38]
2025-02-05T04:51:12.147923549Z #
2025-02-05T04:51:12.147935632Z # A fatal error has been detected by the Java Runtime Environment:
2025-02-05T04:51:12.147936757Z #
2025-02-05T04:51:12.147937465Z #  SIGILL (0x4) at pc=0x0000ee7f37d40c5c, pid=1, tid=7
2025-02-05T04:51:12.147938382Z #
2025-02-05T04:51:12.147939090Z # JRE version:  (21.0.5+11) (build )
2025-02-05T04:51:12.147939923Z # Java VM: OpenJDK 64-Bit Server VM (21.0.5+11-Ubuntu-1ubuntu124.10, mixed mode, sharing, tiered, compressed oops, compressed class ptrs, g1 gc, linux-aarch64)
2025-02-05T04:51:12.147945631Z # Problematic frame:
2025-02-05T04:51:12.147946590Z # j  java.lang.System.registerNatives()V+0 [email protected]
2025-02-05T04:51:12.147947048Z #
2025-02-05T04:51:12.147947506Z # Core dump will be written. Default location: Core dumps may be processed with "/usr/share/apport/apport -p%p -s%s -c%c -d%d -P%P -u%u -g%g -- %E" (or dumping to //core.1)
2025-02-05T04:51:12.147948090Z #
2025-02-05T04:51:12.147950756Z # An error report file with more information is saved as:
2025-02-05T04:51:12.147965714Z # /tmp/hs_err_pid1.log
2025-02-05T04:51:12.150573964Z [0.010s][warning][os] Loading hsdis library failed
2025-02-05T04:51:12.153132590Z #
2025-02-05T04:51:12.153135423Z # The crash happened outside the Java Virtual Machine in native code.
2025-02-05T04:51:12.153136548Z # See problematic frame for where to report the bug.
2025-02-05T04:51:12.153137340Z #
2025-02-05T04:51:12.153173922Z 
2025-02-05T04:51:12.153176172Z [error occurred during error reporting (), id 0x5, SIGTRAP (0x5) at pc=0x0000ee7f506f1b38]
2025-02-05T04:52:12.254493389Z #
2025-02-05T04:52:12.254553134Z # A fatal error has been detected by the Java Runtime Environment:
2025-02-05T04:52:12.254557508Z #
2025-02-05T04:52:12.254558591Z #  SIGILL (0x4) at pc=0x0000ed7c9fd40c5c, pid=1, tid=7
2025-02-05T04:52:12.254561966Z #
2025-02-05T04:52:12.254574673Z # JRE version:  (21.0.5+11) (build )
2025-02-05T04:52:12.254591464Z # Java VM: OpenJDK 64-Bit Server VM (21.0.5+11-Ubuntu-1ubuntu124.10, mixed mode, sharing, tiered, compressed oops, compressed class ptrs, g1 gc, linux-aarch64)
2025-02-05T04:52:12.254593672Z # Problematic frame:
2025-02-05T04:52:12.254644626Z # j  java.lang.System.registerNatives()V+0 [email protected]
2025-02-05T04:52:12.254665874Z #
2025-02-05T04:52:12.254667457Z # Core dump will be written. Default location: Core dumps may be processed with "/usr/share/apport/apport -p%p -s%s -c%c -d%d -P%P -u%u -g%g -- %E" (or dumping to //core.1)
2025-02-05T04:52:12.254668582Z #
2025-02-05T04:52:12.254728702Z # An error report file with more information is saved as:
2025-02-05T04:52:12.254743909Z # /tmp/hs_err_pid1.log
2025-02-05T04:52:12.257321274Z [0.010s][warning][os] Loading hsdis library failed
2025-02-05T04:52:12.259905263Z #
2025-02-05T04:52:12.259914429Z # The crash happened outside the Java Virtual Machine in native code.
2025-02-05T04:52:12.259915679Z # See problematic frame for where to report the bug.
2025-02-05T04:52:12.259944010Z #
2025-02-05T04:52:12.260000297Z 
2025-02-05T04:52:12.260029878Z [error occurred during error reporting (), id 0x5, SIGTRAP (0x5) at pc=0x0000ed7cb81c1b38]

When I migrate this VM to an M1 CPU host with otherwise identical configuration, it works fine:

# docker container list
CONTAINER ID   IMAGE                          COMMAND                  CREATED        STATUS                  PORTS                                                                      NAMES
de5069340ed6   owncloud/ocis-rolling:latest   "/bin/sh -c 'ocis co…"   45 hours ago   Up 29 hours             9200/tcp                                                                   ocis_full-collaboration-1
e1f68dfb7e85   owncloud/ocis-rolling:latest   "/bin/sh -c 'ocis in…"   45 hours ago   Up 29 hours             9200/tcp                                                                   ocis_full-ocis-1
c518a5d51e74   apache/tika:latest-full        "/bin/sh -c 'exec ja…"   45 hours ago   Up 29 hours             9998/tcp                                                                   ocis_full-tika-1
9582b0d44329   collabora/code:24.04.10.2.1    "bash -c 'coolconfig…"   45 hours ago   Up 29 hours (healthy)   9980/tcp                                                                   ocis_full-collabora-1
2224657e4b19   traefik:v3.2.2                 "/entrypoint.sh --lo…"   45 hours ago   Up 29 hours             0.0.0.0:80->80/tcp, :::80->80/tcp, 0.0.0.0:443->443/tcp, :::443->443/tcp   ocis_full-traefik-1

Any ideas how to rectify?

[prohtex]

This seems to be an M4-specific issue:

corretto/corretto-21#85

https://bugs.openjdk.org/browse/JDK-8345296

[prohtex]

Per this link, the fix seems to be JAVA_OPTS=“-XX:UseSVE=0” but I'm not sure where to put that...

https://forums.docker.com/t/image-builds-fail-on-new-macbook-despite-working-fine-on-prior-apple-silicon/145772/5

@micbar @ScharfViktor please help!

[saw-jan]

Per this link, the fix seems to be JAVA_OPTS=“-XX:UseSVE=0” but I'm not sure where to put that...

This might be in the tika service:

ocis/deployments/examples/ocis_full/tika.yml

Lines 3 to 10 in 1f92622

	tika:
	image: ${TIKA_IMAGE:-apache/tika:latest-full}
	# release notes: https://tika.apache.org
	networks:
	ocis-net:
	restart: always
	logging:
	driver: ${LOG_DRIVER:-local}

You can check if this helps:

  tika:
    image: ${TIKA_IMAGE:-apache/tika:latest-full}
    # release notes: https://tika.apache.org
    networks:
      ocis-net:
    restart: always
    environment:
      JAVA_OPTS: "-XX:UseSVE=0"
    logging:
      driver: ${LOG_DRIVER:-local}

[saw-jan]

CC @kobergj @2403905

[prohtex]

tika:
image: ${TIKA_IMAGE:-apache/tika:latest-full}
# release notes: https://tika.apache.org
networks:
ocis-net:
restart: always
environment:
JAVA_OPTS: "-XX:UseSVE=0"
logging:
driver: ${LOG_DRIVER:-local}

Hi @saw-jan! Tried that, but no luck so far. Also tried adding to .env. I thought maybe something needed to be compiled in, like an ldflag?

[wkloucek]

oCIS does not use JAVA.

3eaa53f68e32 apache/tika:latest-full "/bin/sh -c 'exec ja…" 13 minutes ago Restarting (127) 32 seconds ago

You can see that it's Tika. This is not an oCIS bug.

I think it would be a better fit for the Tika issue tracker: https://issues.apache.org/jira/projects/TIKA/issues/

[wkloucek]

Per this link, the fix seems to be JAVA_OPTS=“-XX:UseSVE=0” but I'm not sure where to put that...

see also eg. https://stackoverflow.com/questions/5241743/how-do-i-use-the-java-opts-environment-variable

This means it needs to go in the docker compose environment section of the tika service, see also https://docs.docker.com/compose/how-tos/environment-variables/set-environment-variables/

[prohtex]

Per this link, the fix seems to be JAVA_OPTS=“-XX:UseSVE=0” but I'm not sure where to put that...

see also eg. https://stackoverflow.com/questions/5241743/how-do-i-use-the-java-opts-environment-variable

This means it needs to go in the docker compose environment section of the tika service, see also https://docs.docker.com/compose/how-tos/environment-variables/set-environment-variables/

Thank you @wkloucek, I will log the issue with the Tika team. I added the env var per @saw-jan suggestion, but it didn't work. Is there somewhere else in the container it needs to be set?

I tried:

-adding to .env as JAVA_OPTS="-XX:UseSVE=0"
-adding to tika.yml as JAVA_OPTS: -XX:UseSVE=0
-adding to tika.yml as JAVA_OPTS: "-XX:UseSVE=0"

[saw-jan]

Maybe try JAVA_TOOL_OPTIONS: "-XX:UseSVE=0"?
Suggested in https://stackoverflow.com/questions/5241743/how-do-i-use-the-java-opts-environment-variable

[prohtex]

Maybe try JAVA_TOOL_OPTIONS: "-XX:UseSVE=0"? Suggested in https://stackoverflow.com/questions/5241743/how-do-i-use-the-java-opts-environment-variable

@saw-jan you are my hero! That did it :)

[prohtex]

@wkloucek I will log my issue with Tika but in the meantime, it seems adding this to the ocis_full deployment example would be a good idea. Thanks everyone.