Re: [Wireshark-dev] Packet Loss due to Disk Contention with Running Dumpcap in a high packet rate environment
Hi Kurt, I have purchased a 300 GB SSD and a PCI-e Disk Controller. I am still in the planning stage as to how to set this up. My initial ideas are to: - Leave the SSD on the motherboard Disk Controller - Connect the Platter Hard Drives to the PCI-e Disk Controller - Send the initial capture files to be stored on the SSD - Use RSync to move the files from the SSD to the Platter Hard Drives ( --remove-source-files ) (this is required as I am receiving 1.5 TB a day and the SSD is only 300 GB) ( http://www.cyberciti.biz/tips/linux-use-rsync-transfer-mirror-files-directories.html ) ( http://www.cyberciti.biz/faq/linux-unix-bsd-appleosx-rsync-delete-file-after-transfer/ ) ( http://unix.stackexchange.com/questions/12198/preserve-the-permissions-with-rsync ) ( http://stackoverflow.com/questions/9586798/rsync-over-ssh-preserve-ownership-only-for-www-data-owned-files ) (http://www.comentum.com/rsync.html) - Investigate limiting the Bandwidth of RSync to reduce contention on the Platter Drives if required. (http://www.cyberciti.biz/faq/throttle-disk-io-rate-limit-disk-io/) Thoughts? -John On Wed, Jan 9, 2013 at 6:48 PM, ws.dev.l...@nospam.knochner.com wrote: On 2012-12-14 14:13, John Powell wrote: Thanks for the heads up - I have obtained a 300 G SSD that I will experiment with. I will let you know if I find the same result. I'm just curious. What are your results with the SSD? BTW: did you try to 'ionice' dumpcap? Cheers Kurt __**__** ___ Sent via:Wireshark-dev mailing list wireshark-dev@wireshark.org Archives: http://www.wireshark.org/**lists/wireshark-devhttp://www.wireshark.org/lists/wireshark-dev Unsubscribe: https://wireshark.org/mailman/**options/wireshark-devhttps://wireshark.org/mailman/options/wireshark-dev mailto:wireshark-dev-request@**wireshark.orgwireshark-dev-requ...@wireshark.org ?subject=**unsubscribe ___ Sent via:Wireshark-dev mailing list wireshark-dev@wireshark.org Archives:http://www.wireshark.org/lists/wireshark-dev Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev mailto:wireshark-dev-requ...@wireshark.org?subject=unsubscribe
Re: [Wireshark-dev] SCTP De-chunking support
On Jan 10, 2013, at 5:07 PM, vineeth vijay wrote: Hi All, Has the de-chunking of SCTP within wireshark been attempted yet? I noticed some old conversations in mailing list in this regard, but nothing concrete has turned up yet. While trying to do this in tshark, I have tried calling tshark's process_packet() function from packet-sctp.c file in dissectors but got nowhere due to linking issues. I feel the way to do this would be: 1) Create a global copy of entire frame at initial stage (Is there any other way to access the entire frame structure from packet-sctp where ultimately the decision whether to do de-chunking or not would be made. ) 2) In case there are several chunks in the packet, allow the completion of processing till first chunk and create composite tvbs consisting of eth+ip+sctp_header+remaining_individual_chunks. 3) Correct IP checksums and length in the composite Tvb. 4) Process these tvb's individuallly. (Is this possible with the rule to have a single capture file at a time?? Can a capture file structure be modified on the fly?) Is the above process doable without breaking wireshark/tshark processing structure? Can anybody suggest a better solution... Hi, what problem are you trying to solve? Wireshark supports dissecting the upper layer paylaod for bundled DATA chunks for ages... Best regards Michael Vineeth ___ Sent via:Wireshark-dev mailing list wireshark-dev@wireshark.org Archives:http://www.wireshark.org/lists/wireshark-dev Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev mailto:wireshark-dev-requ...@wireshark.org?subject=unsubscribe ___ Sent via:Wireshark-dev mailing list wireshark-dev@wireshark.org Archives:http://www.wireshark.org/lists/wireshark-dev Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev mailto:wireshark-dev-requ...@wireshark.org?subject=unsubscribe
Re: [Wireshark-dev] SCTP De-chunking support
On Jan 10, 2013, at 8:49 PM, vineeth vijay wrote: Hi, Dissection is fine. What I was wondering is whether it is possible to show these individual data chunks as separate frames themselves. But they are in the same frame. I really prefer not to show them in a way they have not been on the wire. Basically agreed on the above point. Changing the default behavior may not be good due to all the copied lower layer bytes and resulting increase in the size of capture in case there are 4-5 chunks per packet. But still feel it would be a nice optional feature to have when doing actual offline analysis. I do understand that it is sometimes hard to find the application layer packet when using display filters and there are multiple application layer packets bundled in a single frame. I also have traces with a large number of bundled chunks. Hence, when i apply display filter , only the chunks with exact matches should be visible. Is this supported currently? No. Filtering is based on packets. Not sure how to improve that. We can't show 'half' of a packet. However, there might be ways to draw your attention to the upper layer packet which matches the filter. Regarding above point, would like to suggest that the packet information being displayed can be restricted to the PDU which actually matches the display filter. E.g out of an SCTP packet carrying 3-4 M3UA chunks, the pinfo of only the chunk matching the filter can be displayed? Thinking about this... What about displaying only the frames, which match a display filter (like today). However, it might be helpful to highlight that part (like the M3UA packet) which matches the display filter. This should allow to find the upper layer packet pretty fast. What do you think? Best regards Michael Vineeth On Fri, Jan 11, 2013 at 12:54 AM, Michael Tuexen michael.tue...@lurchi.franken.de wrote: On Jan 10, 2013, at 5:31 PM, vineeth vijay wrote: Hi, Dissection is fine. What I was wondering is whether it is possible to show these individual data chunks as separate frames themselves. But they are in the same frame. I really prefer not to show them in a way they have not been on the wire. Hence, when i apply display filter , only the chunks with exact matches should be visible. Is this supported currently? No. Filtering is based on packets. Not sure how to improve that. We can't show 'half' of a packet. However, there might be ways to draw your attention to the upper layer packet which matches the filter. Best regards Michael Currently , i use the below tool for this purpose: http://frox25.no-ip.org/~mtve/wiki/SctpDechunk.html Regards, Vineeth what problem are you trying to solve? Wireshark supports dissecting the upper layer paylaod for bundled DATA chunks for ages... Best regards Michael Vineeth ___ Sent via:Wireshark-dev mailing list wireshark-dev@wireshark.org Archives:http://www.wireshark.org/lists/wireshark-dev Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev mailto:wireshark-dev-requ...@wireshark.org?subject=unsubscribe ___ Sent via:Wireshark-dev mailing list wireshark-dev@wireshark.org Archives:http://www.wireshark.org/lists/wireshark-dev Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev mailto:wireshark-dev-requ...@wireshark.org?subject=unsubscribe ___ Sent via:Wireshark-dev mailing list wireshark-dev@wireshark.org Archives:http://www.wireshark.org/lists/wireshark-dev Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev mailto:wireshark-dev-requ...@wireshark.org?subject=unsubscribe ___ Sent via:Wireshark-dev mailing list wireshark-dev@wireshark.org Archives:http://www.wireshark.org/lists/wireshark-dev Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev mailto:wireshark-dev-requ...@wireshark.org?subject=unsubscribe ___ Sent via:Wireshark-dev mailing list wireshark-dev@wireshark.org Archives:http://www.wireshark.org/lists/wireshark-dev Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev mailto:wireshark-dev-requ...@wireshark.org?subject=unsubscribe ___ Sent via:Wireshark-dev mailing list wireshark-dev@wireshark.org Archives:http://www.wireshark.org/lists/wireshark-dev Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev
Re: [Wireshark-dev] SCTP De-chunking support
Hi, Yes, highlighting would work too. Ultimately the application info corresponding to display filter should be visible easily without the need to scroll through the entire frame. Any suggestions on how to achieve this? I think GUI coloring implementation would paint the entire frame with the same color,wouldn't it? Vineeth On Fri, Jan 11, 2013 at 1:44 AM, Michael Tuexen michael.tue...@lurchi.franken.de wrote: On Jan 10, 2013, at 8:49 PM, vineeth vijay wrote: Hi, Dissection is fine. What I was wondering is whether it is possible to show these individual data chunks as separate frames themselves. But they are in the same frame. I really prefer not to show them in a way they have not been on the wire. Basically agreed on the above point. Changing the default behavior may not be good due to all the copied lower layer bytes and resulting increase in the size of capture in case there are 4-5 chunks per packet. But still feel it would be a nice optional feature to have when doing actual offline analysis. I do understand that it is sometimes hard to find the application layer packet when using display filters and there are multiple application layer packets bundled in a single frame. I also have traces with a large number of bundled chunks. Hence, when i apply display filter , only the chunks with exact matches should be visible. Is this supported currently? No. Filtering is based on packets. Not sure how to improve that. We can't show 'half' of a packet. However, there might be ways to draw your attention to the upper layer packet which matches the filter. Regarding above point, would like to suggest that the packet information being displayed can be restricted to the PDU which actually matches the display filter. E.g out of an SCTP packet carrying 3-4 M3UA chunks, the pinfo of only the chunk matching the filter can be displayed? Thinking about this... What about displaying only the frames, which match a display filter (like today). However, it might be helpful to highlight that part (like the M3UA packet) which matches the display filter. This should allow to find the upper layer packet pretty fast. What do you think? Best regards Michael Vineeth On Fri, Jan 11, 2013 at 12:54 AM, Michael Tuexen michael.tue...@lurchi.franken.de wrote: On Jan 10, 2013, at 5:31 PM, vineeth vijay wrote: Hi, Dissection is fine. What I was wondering is whether it is possible to show these individual data chunks as separate frames themselves. But they are in the same frame. I really prefer not to show them in a way they have not been on the wire. Hence, when i apply display filter , only the chunks with exact matches should be visible. Is this supported currently? No. Filtering is based on packets. Not sure how to improve that. We can't show 'half' of a packet. However, there might be ways to draw your attention to the upper layer packet which matches the filter. Best regards Michael Currently , i use the below tool for this purpose: http://frox25.no-ip.org/~mtve/wiki/SctpDechunk.html Regards, Vineeth what problem are you trying to solve? Wireshark supports dissecting the upper layer paylaod for bundled DATA chunks for ages... Best regards Michael Vineeth ___ Sent via:Wireshark-dev mailing list wireshark-dev@wireshark.org Archives:http://www.wireshark.org/lists/wireshark-dev Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev mailto:wireshark-dev-requ...@wireshark.org ?subject=unsubscribe ___ Sent via:Wireshark-dev mailing list wireshark-dev@wireshark.org Archives:http://www.wireshark.org/lists/wireshark-dev Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev mailto:wireshark-dev-requ...@wireshark.org ?subject=unsubscribe ___ Sent via:Wireshark-dev mailing list wireshark-dev@wireshark.org Archives:http://www.wireshark.org/lists/wireshark-dev Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev mailto:wireshark-dev-requ...@wireshark.org ?subject=unsubscribe ___ Sent via:Wireshark-dev mailing list wireshark-dev@wireshark.org Archives:http://www.wireshark.org/lists/wireshark-dev Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev mailto:wireshark-dev-requ...@wireshark.org ?subject=unsubscribe ___ Sent via:Wireshark-dev mailing list wireshark-dev@wireshark.org Archives:
Re: [Wireshark-dev] SCTP De-chunking support
Hi, I understood the idea. It would help in easier detection of the relevant upper layer info in large packets. What i would like to know is how it could be implemented. Setting some sort of flag for the filter specific chunk bytes, so that GUI/GTK colors it differently? Sorry, but i am not much familiar with GTK. Vineeth On Fri, Jan 11, 2013 at 4:08 AM, Michael Tuexen michael.tue...@lurchi.franken.de wrote: On Jan 10, 2013, at 9:44 PM, vineeth vijay wrote: Hi, Yes, highlighting would work too. Ultimately the application info corresponding to display filter should be visible easily without the need to scroll through the entire frame. Any suggestions on how to achieve this? I think GUI coloring implementation would paint the entire frame with the same color,wouldn't it? No, what I mean is the following: Assume you have an SCTP packet with 5 DATA chunks each containing an M3UA message. The packet is shown because you filtered for a field in the third M3UA message. Then only the third M3UA part would be colored specifically. The rest of the packet is shown, but not in this color. Do you get the idea from my description? Would that address your issue? Best regards Michael Vineeth On Fri, Jan 11, 2013 at 1:44 AM, Michael Tuexen michael.tue...@lurchi.franken.de wrote: On Jan 10, 2013, at 8:49 PM, vineeth vijay wrote: Hi, Dissection is fine. What I was wondering is whether it is possible to show these individual data chunks as separate frames themselves. But they are in the same frame. I really prefer not to show them in a way they have not been on the wire. Basically agreed on the above point. Changing the default behavior may not be good due to all the copied lower layer bytes and resulting increase in the size of capture in case there are 4-5 chunks per packet. But still feel it would be a nice optional feature to have when doing actual offline analysis. I do understand that it is sometimes hard to find the application layer packet when using display filters and there are multiple application layer packets bundled in a single frame. I also have traces with a large number of bundled chunks. Hence, when i apply display filter , only the chunks with exact matches should be visible. Is this supported currently? No. Filtering is based on packets. Not sure how to improve that. We can't show 'half' of a packet. However, there might be ways to draw your attention to the upper layer packet which matches the filter. Regarding above point, would like to suggest that the packet information being displayed can be restricted to the PDU which actually matches the display filter. E.g out of an SCTP packet carrying 3-4 M3UA chunks, the pinfo of only the chunk matching the filter can be displayed? Thinking about this... What about displaying only the frames, which match a display filter (like today). However, it might be helpful to highlight that part (like the M3UA packet) which matches the display filter. This should allow to find the upper layer packet pretty fast. What do you think? Best regards Michael Vineeth On Fri, Jan 11, 2013 at 12:54 AM, Michael Tuexen michael.tue...@lurchi.franken.de wrote: On Jan 10, 2013, at 5:31 PM, vineeth vijay wrote: Hi, Dissection is fine. What I was wondering is whether it is possible to show these individual data chunks as separate frames themselves. But they are in the same frame. I really prefer not to show them in a way they have not been on the wire. Hence, when i apply display filter , only the chunks with exact matches should be visible. Is this supported currently? No. Filtering is based on packets. Not sure how to improve that. We can't show 'half' of a packet. However, there might be ways to draw your attention to the upper layer packet which matches the filter. Best regards Michael Currently , i use the below tool for this purpose: http://frox25.no-ip.org/~mtve/wiki/SctpDechunk.html Regards, Vineeth what problem are you trying to solve? Wireshark supports dissecting the upper layer paylaod for bundled DATA chunks for ages... Best regards Michael Vineeth ___ Sent via:Wireshark-dev mailing list wireshark-dev@wireshark.org Archives:http://www.wireshark.org/lists/wireshark-dev Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev mailto:wireshark-dev-requ...@wireshark.org ?subject=unsubscribe ___ Sent via:Wireshark-dev mailing list wireshark-dev@wireshark.org Archives:http://www.wireshark.org/lists/wireshark-dev Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev