|
|
Special Considerations for Release 2.0.7 ² Other Special Considerations
This chapter describes special considerations that apply to Release 2.0.7 of the LightStream 2020 enterprise ATM switch. These special considerations are drawn from reports of hands-on experience with the product by LightStream developers and LightStream SQA personnel.
Most of the special considerations here are provided with a LightStream Problem Report Number (PRN). If you contact Cisco Systems about any of these special considerations, please refer to them by PRN.
A terminal may lose carriage return or echo functionality.
Workaround: Execute the bash tset command to restore the lost property.
(LIGle00221) The primary TCS hub and the primary (active) switch fabric may be on different switch cards in a redundant system. This has caused confusion when answering the question, which card is primary.
The CLI show chassis primaryswitch command indicates which switch fabric is active. The primary TCS hub is indicated in any of the following three ways:
Step 1 The following TCS command:
Step 2 The use of uppercase letters in the TCS HUB prompt. That is, the prompt TCS HUB <<B>> indicates that the TCS hub on the card in slot B is active, and the prompt tcs hub <<b>> indicates that it is secondary. (In both cases, the name of the hub card to which you are connected is indicated by the letter in brackets.)
Step 3 The LED labelled TCS SEL on the switch card is lit.
On systems with at least one Release 1 switch card (SC1), changing which switch is primary with the CLI command set chassis primaryswitch causes the chassis to reboot.
(LIGle00225) The configuration tool and CLI allow rates to be set to 6Mbits/s aggregate for low speed cards. The maximum throughput per port for the low speed card is about 4.1 Mbits/sec and 4600 frames/sec, using committed traffic and an 88-byte frame. If the rate is set above this limit, the excess may be dropped.
Workaround: Do not set rates higher than 3.84 Mbits/sec per port on low-speed cards.
Although limited functionality for a frame relay NNI interface is present in the Release 2.0.7 software, this is not a supported facility.
Looping of a Frame Relay UNI interface provides no useful diagnostic information because the UNI protocol is asymmetric. LightStream supports internal looping of such a port by first converting it to an NNI interface. A successful loop sets the administrative state of the port to testing and the operational state to up.
In Release 2.0.7, FDDI and Ethernet ports cannot be looped.
The CLI utility allows you to directly modify variables in the MIB using the setsnmp command. This command performs no validation of its input data and does not prevent you from setting a MIB variable to an inappropriate value.
Workaround: The preferred approach is to use the following command within the CLI to change MIB variables:
set <object-type> [object-id] [parameter1...]
This approach gives the CLI enough context to validate the new value for the variable before performing the change to the MIB.
The POST timeout value on all packet line cards and cell line cards (PLC and CLC) must be modified to support new PLC and CLC POST.
Verify the POST timeout value of cards in your system by typing the following at the TCS command prompt:
read slot ee 44 word
The value read should be 0x0e4e.
If this is incorrect, enter the following at the TCS command prompt:
write slot ee 44 e4e word
The oc3InfoMediumType object can be set to 1 for SONET operation or to 2 for SDH operation of the OC-3 card.
The default for OC-3 ports is to use SONET. To set a series of OC-3 ports to use SDH instead of SONET, carry out the following procedures:
Identify all OC-3 ports in your network that you wish to be running SDH.
This procedure must be done for each switch in turn.
Step 1 Log in to the switch.
Step 2 Start the CLI
Step 3 Enter protected mode with the protected command and password, as follows:
pro
Step 4 Set the SNMP community name to a community that allows writing data. In the following example command, this community is named write:
set snmp community write
Step 5 Set the configuration lock so that your changes are written to the configuration database, as follows:
set config lock
Step 6 Configure identified OC3 ports to use SDH, as follows:
setsnmp oc3InfoMediumType.cc0pp 2
The MIB object name must be typed with uppercase letters exactly as shown, e.g. oc3InfoMediumType.10003 for port 3 on OC-3 card 10.
Step 7 When you have reconfigured all the identified OC-3 ports, unlock the configuration database, as follows:
set config unlock
To configure a port for SONET, carry out this same procedure, substituting the following command in Step 6:
setsnmp oc3InfoMediumType.cc0pp 1
Every time you install a card in a slot of the system, you must carry out the following procedure.
Step 1 Log in to the active NP as root.
Step 2 To see the current flash checksums of all cards in the chassis, enter the following command:
sysver -a | more
The correct checksums are follows:
Step 3 If a line card or NP flash checksum is not correct, load it using the following command:
fcload -s slot# -flash -force
Here, slot# represents the slot number of the card with the incorrect flash checksum.
If a new switch card has an incorrect flash checksum, load the flash by following the procedure for Updating the Switch Card Flash Memory.
Step 4 Verify flash again (Step 1 and Step 2 above).
Step 5 If the card still has an incorrect flash checksum after loading flash, contact LightStream Customer Support.
The following sections detail additional special considerations that have been formulated for print late in the product release cycle.
The special considerations in this section are provided with a LightStream identification number. If you contact LightStream Corporation about any of these special considerations, please refer to the identification number.
CLI does not allow setting of bandwidth parameters for virtual circuits to defaults.
Use default values specified in the LightStream 2020 Configuration Guide.
Intermittent port shutdowns on LS1.
During throughput testing on the low-speed card in edge mode at near maximum capacity (over 3 Mbps), the port under test was observed to shut off (drop the RTS signal) for 2 to 3 seconds. In each case, RTS returned and traffic resumed. No action is required.
Deleting PVC hangs up MMA when thousands of PVCs are used.
If hundreds of PVCs are created on a single port, then deleting a PVC can take a long time. The result will be an SNMP timeout to the Net Management tool being used to delete the PVC. The delay could be several minutes when over 100 PVCs are created on a given port.
A bug in HP OpenView limits collections to < 20 variables.
Data collection fails with collections set up for more than 20 variables. This is because HP OpenView formats the request into a buffer which, in this case, exceeds the SNMP standard. In compliance with the SNMP standard, the LightStream software drops such requests.
Workaround: When setting up data collection using HP OpenView, limit the collection to less than 20 variables.
The tar utility is usable with floppy disk but generates errors.
When attempting tar to the floppy drive (for example tar xvf /dev/sd1) when no diskette is present in the drive, various uninformative error messages appear on the console. Similar symptoms occur when using a disk that has not been formatted for an IBM PS/2 (models 50, 60, 70, 80 and compatibles).
Furthermore, when using a floppy disk with media errors, the software process using the device may hang, preventing future access to the floppy or to the terminal where the tar command was being run.
Workaround: Reboot the network processor.
Bandwidth allocation does not check egress port capacity.
Release 1 medium-speed card counts receive errors when a cable is not attached.
When the receive port on a Release 1 medium-speed card is left unattached to a cable or attached to an unterminated cable, the receive error counts will be incremented (if that port is configured, high error- rate traps appear on the console).
The ckswinstall utility behaves poorly when applied to update distributions
The ckswinstall command (to verify a software installation) has several confusing aspects.
Step 1 The ckswinstall utility reports spurious permission mismatches on directories under /usr/app.
Step 2 When using ckswinstall to verify an update distribution (for instance 2.0.7), ckswinstall only verifies the software in the update. On a system running an update, most of the software in use is from the original release, so ckswinstall should also be run on that original release (for example, 2.0.3). When running ckswinstall on the underlying release, spurious errors are reported against software that was upgraded as part of the update.
No procedure is described for checking software on the standby NP.
When verifying a software installation using the ckswinstall utility on a redundant-NP system, the software installation on each of the two NPs must be checked explicitly.
Step 1 Run ckswinstall on the current primary NP.
Step 2 Run ckswinstall on the current backup NP.
rsh other-np ckswinstall...
Periodic rdist to the secondary NP doesn't check for disk space available.
On a redundant NP system, an automatic mechanism updates software and configuration information from the current primary NP to the current secondary NP. If the current secondary NP runs out of disk space, the secondary NP is not kept consistent with the primary NP. Furthermore, no error is reported.
Workaround: Establish a regular maintenance procedure of checking disk space availability on both NPs.
Insertion of disk tray assembly can cause damage to connectors.
Rapid insertion of the disk tray assembly can defeat the guiding mechanism and may result in damage to the pins of the connectors. Please use care when inserting or removing the disk tray.
Switch Fault LED lights when a redundant power supply is inserted.
The fault light on the switch card is controlled by the TCS HUB code which runs on the switch card. To monitor the power supply, the HUB code checks each power supply's status every 750 ms by looking at two signals generated by each power tray:
| Signal | Meaning |
|---|---|
| Type Code | Indicates power tray is present and what type of power supply it is. |
| DC OK | Indicates that the power supply is producing 48 Volts. |
A fault LED occurs when the Type Code indicates that a power tray is present while the DC OK signal indicates that it is not producing 48 volts. The HUB code will also set a bit in the switch card status 1 indicating that a power supply error condition exists.
The fault LED is turned off by the HUB code when the NPC reads the HUBS status register. The writing of the status register by the HUB code and the reading of this register by the NPC are asynchronous events. The NPC will read the status register a minimum of every six seconds.
The status of the two signals immediately following the insertion of a power supply is as follows:
| Signal | Value |
|---|---|
| Type Code | Power supply present. |
| DC OK | False. |
The DC OK signal reads False because it takes a maximum of 750 ms for the power supply to produce 48 Volts after AC power is applied.
If the TCS HUB code checks the status of these two signals during this brief period of time, it considers it an error condition (DC OK = FALSE) and turns on the fault LED.
In these circumstances, ignore the fault LED. Normal Operation is never interrupted.
Shouldn't need to enable NP after power-off/power-on sequence (Release 1 switch cards only).
The proper sequence to bring up a card from the TCS HUB after a power slot off command is to issue the reset slot command. This turns on the card power and initiates the card boot sequence.
Issuing a power slot on command only turns the card VCC on, but does not initiate the card boot sequence.
May be unable to free up sufficient disk space for new release.
When attempting to install a new release, you may be told that insufficient disk space is available on the sd0c partition. Follow the instructions in the release note for deleting old releases to free up sufficient space for the new release. If, having followed these instructions, the installation procedure still reports that insufficient disk space is available on sd0c, contact your support organization for assistance in proceeding.
HUB reported POST PASSED and POST error at the same time.
There are several things which can cause this to happen. The most common is when the TCS power off command is used followed by a TCS power on command.
If you are using TCS commands, the proper method for bringing a card back to life is to use the power off command and then to use the reset command.
The following TCS hub command reports POST passed:
show slot# post
The following TCS hub command reports POST_ERR:
status slot#
The command show chassis listtrunks shows same trunk twice
The CLI command show chassis listtrunks shows a separate record for each trunk from each end. This might be improved in the future and information consolidated into one record per trunk.
Discarded cells are not accounted for in CLI statistics.
On the OC3 card, the per-port Discarded Rcvd Packets statistic in the CLI show port display does not include permit drops. Permit blocks are displayed correctly in the per-VCI statistics.
PVC has no "Save As" file menu item
The PVC and VLI configurators do not support "Save As" function. You can create copies of the database by exiting the PVC / VLI configuration application and copying the.dat/.pag files from a UNIX shell.
Cannot connect to backup NP in diagnostic mode.
Network processor (NP) diagnostics must be run from the hub. The mechanism by which diagnostics are run on some card using the CLI on the primary NP (the CLI test command) is not supported for running diagnostics on a second NP.
Deleted PVC cannot be added in again: "Waiting to be deleted from switch".
The current implementation restriction requires that, after initiating a download, the user must wait until the download completes before trying to modify the records involved in the download. Further, the user needs to reread the database after the download completes before modifying the records.
Fallback from swchgver does not copy config files back.
LightStream configuration information is stored on disk as part of a specific release. Configuration information from an old release is copied forward to a new release as part of installation of a new release. Configuration information, however, is not automatically copied between releases when swchgver is used to change the current release.
For instance, if a node is upgraded to run new software, configuration changes are made, and then the node is downgraded to a previous release, the node will be running with the configuration information that was cached at the time of the upgrade, which does not include the subsequent changes.
After falling back to an old version, the current configuration cache can be downloaded to the node to set the newest configuration.
Alternatively, before falling back to an old version, the files in /usr/app/base-newrel/config can be copied to /usr/app/base- oldrel/config before the fallback (replace newrel and oldrel with the appropriate release numbers, such as 2.0.7 and 2.0.3).
External bridge loops cause loss of VLI functionality.
When configuring Lightstream nodes with external topology loops, configure the bridge priority for the nodes such that one of them will become the spanning tree root.
Rate statistics for ports are inaccurate.
The Rate information produced by the CLI when displaying port statistics is not accurate.
An ATM UNI or FF Port reports Operationally up when disconnected.
Currently the FF and ATM-UNI interfaces use no protocol to provide a link-level reliability indication such as the LMI for FR or Trunk-Up-Down (TUD) for LightStream trunks. The Operational Status on these ports reflects the Administrative Status of the port. The operational status does not indicate physical layer or data link layer status.
Request a way to abort unwanted download.
No function is currently provided to abort long downloads. You might be able to abort downloads using the UNIX kill command, after finding the process ID of the cfg_a process. However, the results could be unpredictable, because some part of the configuration might have been downloaded and some not.
FR should be upgraded to support T1.617a-1993.
The Frame Relay LMI conforms to T1.617-1991 but has not been upgraded to meet T1.617a-1993. This should not affect normal operation of the port, but may affect conformance tests designed to verify the updated standard. The FR port interoperates with LMIs meeting either version of this standard.
The file open/save dialog box filter does not work.
If you are making a backup of the global database using the Save As function, the backup file should end in the.pag extension.
CDN #1745 (NP1 diagnostic fails tests 22,23,24 intermittently).
This is a problem with the revision of the PIT timer used on some NP cards. A modified software algorithm has been put in the NP to make this test more reliable. Some problems remain.
The procedure to make backup copies of install diskettes should be documented in the Operations Guide. It is documented in the Release Notes.
File->Open defaults to the home directory of the ovw initializer.
When a database is opened from the File->Open menu item, the path defaults to the home directory of the initiator of ovw instead of following LSC_DATABASE.
Workaround: Double-click in Directories to navigate to the /usr/OV/databases/ls directory, and then select the appropriate directory.
The release 1 switch card and the release 2 switch card interoperate with the following modems:
Zoom 9624V
Zoom FXV (FX9624V)
Hayes SmartModem 2400 V.24
Using any other modem may cause unpredictable results.
OC-3 card continuously bounced when connected to SONET MUX.
When connecting an OC3 port to a SONET multiplexor, a clocking type of external should normally be used (rather than the default of internal). Use of internal clocking may prevent the port from operating correctly.
External SNMP devices impact system performance.
The NP software does not throttle excess traffic from external SNMP devices. This can consume CPU resources and can cause timeouts, such as spanning tree timeouts, degrading system performance. Typical sources of SNMP packets include the LightStream monitor, HP Openview statistical graphing facilities displaying a large number of MIB variables, and third-party network management applications. When this happens, you must reduce the SNMP traffic from the external device. You might reduce the number of variables being tracked, increase the polling interval, or reduce the number of concurrent network management processes which are requesting data.
The command set card card# testing no longer works, and is rejected by the CLI. The test command puts the card in the testing state. Use the command test card# -m to put card card# in the testing state, load diagnostics, and establish an I/O tap connection from the CLI to card card#.
The Release 2.0 Installation and Troubleshooting Manual says to search the /etc/services file for the string snmp-traps. It should say to search for the string snmp-trap.
The Release 2.0 Installation and Troubleshooting Manual suggests to change the polling interval to 1 minute. The polling interval should be left at the default of 5 minutes.
|
|