Known Problems Fixed in This Version

Under some stress situations (for example, a port is continuously enabled and then disabled to run LACP), the system panics due to data page fault with the following stack trace:

panic+0x8c
report_trap_or_int_and_panic+0x94
interrupt+0x4e4
ihandler+0x940
lacp_processing_poll_timer+0x128
lacp_ctrl+0x1e8
lacp_poll_timeout+0x60
invoke_callouts_for_self+0x234
soft_intr_handler+0x1
sw_service+0x1c
mp_ext_interrupt+0x478
ihandler+0x91c

When creating a MANUAL mode link aggregate from a set of ports, if some of the ports do not exist, APA does not create the link aggregate.

When a user application sends a DL_HP_MULTICAST_LIST_REQ primitive to a link aggregate or failover group, APA returns the error code 233 (EOPNOTSUPP).

Unable to create a MANUAL mode link aggregate from ports with different Group Capability values.

Unnecessary failover can occur with cost-based failover groups.

No information on proactive failover is saved to the var/adm/syslog/syslog.log file.

The NetTL messages that describe proactive failover events do not provide enough information.

When the LACP timeout values of link partners are changed, APA responds in a manner that does not conform to the IEEE 802.3ad standard. This can cause a temporary connection loss.

If an LACP port's link partner times out and then recovers, the LACP port might not rejoin its original link aggregate.

LACP ports might take up to 90 seconds to detect a unidirectional link failure.

When one link of a LACP link aggregate becomes unidirectional (a link can either receive or transmit), network connectivity of the aggregate is lost.

Sometimes the LACP states of APA does not match those of its link partners. This can cause a temporary connection loss.

When one link partner is not ready to receive packets but others are ready, network connectivity is lost .

APA sets the collecting and distributing bits in the LACP PDU when the synchronization bit is cleared.

Under some stress situations (for example, a port is continuously enabled and then disabled to run LACP), the system panics with data page fault in the lacp_tx_sm() function.

Under some stress situations (for example, a port is continuously enabled and then disabled to run LACP), the system panics with data page fault in the lacp_notify_msg() function.

When a user application sends a DL_HP_HW_RESET_REQ primitive to a link aggregate or failover group, APA returns the error code 233 (EOPNOTSUPP).

Under some stress situations, the system panics with a data page fault in LACP subsystem.

Four ports are connected to two Nortel switches using Split Multi-Link Trunking (SMLT), and LACP is used for trunking. When one of the Nortel switches is powered down and then powered up, the LACP ports migrate to a different link aggregate on the server. Because the new link aggregate has no IP address configured, the network connection is lost completely.

Four ports are connected to two Nortel switches using Split Multi-Link Trunking (SMLT), and LACP is used for trunking. When one of the switches is powered down, the network connection is lost completely. SMLT is a proprietary technology to make link aggregates across switches.

APA removes a port from LACP link aggregates only after the port fails to send an LACP PDU for more than 90 seconds.

When two LACP ports on an HP server are connected to two Nortel switches using SMLT (each port is connected to a unique switch), they initially form a link aggregate. After one switch fails, the link aggregate disappears.

The system might hang after setting ifAdmin states of APA link aggregates. This is a rare occurrence.

The lan*conf commands can log unneeded information to NetTL.

A bad_news panic in the apa_lacp_check_linkagg() function with the following stack trace:

panic_save_regs_switchstack+0x110
panic+0x490
bad_news+0x440
bubbleup+0x8e0
apa_lacp_check_linkagg+0xa0
apa_update_linkagg_properties+0x2c0
apa_ioctl_del_linkagg_ports+0x3b0
apa_ioctl_clear+0x650
apa_process_event+0x460
apa_process_queue+0x330
invoke_callouts_for_self+0x2e0
soft_intr_handler+0x240
external_interrupt+0x620
bubbleup+0x8e0
deadlock_check+0x190
sl_pre_check+0x1f0
spinlockx+0x90
prioceil_restore_pri+0x90
rwlock_unlock+0x430
pm_unlink_thread_from_active+0x50
exit+0x4c0
rexit+0x40
syscall+0x1430

During heavy loads, ports in PAgP link aggregates drop out of link aggregates and then rejoin. This disrupts the traffic flow across the link aggregate interface.

The lanqueryconf command might put ports having VLANs configured on them into failover groups. The lanapplyconf command fails if it detects these ports inside a failover group.

The lanapplyconf command cannot create failover groups whose member ports have an MTU of 259 or less.

The lanapplyconf command dumps core if the NODE_NAME value in the /etc/lanmon/lanconfig.ascii file is different from the server's hostname.

While experiencing a large number of port state changes, HP APA might place LACP ports in the wrong link aggregate.

When ports in a LACP link aggregate experience many protocol events, they might move to the wrong LACP link aggregate.

System hangs after repeatedly starting and stopping PAgP or LACP on ports.