backupList is not planned because the coordinates approach was sufficient —Eric
On May 9, 2017, at 6:57 AM, Ori Finkelman <o...@qwilt.com<mailto:o...@qwilt.com>> wrote: Hi again, I understand now that the "backupList" feature does not exist yet. What is the status of this feature ? is it planned ? Thanks, Ori On Mon, May 8, 2017 at 4:18 PM, Ori Finkelman <o...@qwilt.com<mailto:o...@qwilt.com>> wrote: Hi, Following up on this one, it seems that both czf attributes described in this thread, the "coordinates" and the "backupList" are not documented in the official docs in http://trafficcontrol.incubator.apache.org/docs/latest/admin/traffic_ops_ using.html#the-coverage-zone-file-and-asn-table Is there a plan to update the documentation ? should I open a JIRA for it ? Thanks, Ori On Thu, Mar 30, 2017 at 8:45 PM, Jeff Elsloo <jeff.els...@gmail.com> wrote: Yes, that's correct. -- Thanks, Jeff On Thu, Mar 30, 2017 at 11:20 AM, Eric Friedrich (efriedri) <efrie...@cisco.com> wrote: Thanks Jeff- Could I think of it as the following? Echoing back to be sure I understand... If there is a lat/long for a cache group in the CZF file, any client hit to that CG should use the CZF lat/long as it client’s lat/long instead of using geolocation. For the purposes of finding closest cache group, the client’s location (from CZF as above or from Geolocation provider) will be compared against the location of the cache’s as configuration in Traffic Op’s CG record? —Eric On Mar 30, 2017, at 1:07 PM, Jeff Elsloo <jeff.els...@gmail.com> wrote: It could now be considered the "average" of the location of the clients within that section of the CZF, however, it should be noted that the addition of the geo coordinates to the CZF is relatively new. Previously we never had the ability to specify lat/long on those cachegroups, and we solely relied on those specified in edgeLocations, meaning that the matches had to be 1:1. Adding the coordinates allowed us to cover edge cases and miss scenarios and stick to the CZF whenever possible. Previously when we had no coordinates, and we had a hit in the CZF but not corresponding hit within the edgeLocations (health, assignments, etc), we would fall back to the Geolocation provider. -- Thanks, Jeff On Thu, Mar 30, 2017 at 5:29 AM, John Shen (weifensh) <weife...@cisco.com> wrote: Thanks Jeff and Oren for the discussion. I agree now that lat/long from CZF is the “average” location of clients, and lat/long from Ops is the location of a certain Cache Group. So it appears to be reasonable to use them as source and dest to calculate the distance. Thanks, John On 30/03/2017, 6:55 PM, "Oren Shemesh" <or...@qwilt.com> wrote: Jeff, having read this conversation more than once, I believe there is a misunderstanding regarding the ability to provide coordinates for cache groups both in the CZF and in the TO DB. Here is what I believe is a description which may help understanding the current behaviour: The coordinates specified in the CZF for a cache group are not supposed to be the exactly same as the coordinates in the TO DB for the same cache group. This is because they do not represent the location of the caches of the group. They represent the (average) location of clients found in the subnets specified for this cache group. This, I believe, explains both the behaviour of the code (Why the coordinates from the CZF are used for the source, but the coordinates from the TO DB are used for the various candidate cache groups), and the fact that there is a 'duplication'. Is this description true ? On Wed, Mar 29, 2017 at 7:02 PM, Jeff Elsloo <els...@apache.org> wrote: The cachegroup settings in the Traffic Ops GUI end up in the `edgeLocations` section of the CRConfig. This is the source of truth for where caches are deployed, logically or physically. We do not provide a means to generate a CZF in Traffic Ops, so it's up to the end user to craft one to match what is in Traffic Ops. There are several cases that need to be accounted for where a hit in the CZF does match what's in `edgeLocations`, but cannot be served there due to cache health, delivery service health, or delivery service assignments. The other edge case is a hit where no `edgeLocation` exists, which again, must be accounted for. Presumably we have higher fidelity data in our CZF than we would in our Geolocation provider and we should use it whenever possible. Think about this: what if you use the same CZF for two configured CDNs, but one of the two CDNs only has caches deployed to 50% of the cache groups defined in the CZF. Would we want to use the Geolocation provider in the event that our source address matches a cachegroup that does not have any assigned caches? We would ideally have as much granularity as possible in the CZF, then use that to inform the decision about which cachegroup should service the request instead of falling back to a lower fidelity datasource. This is especially true in the case of RFC 1918 addresses that might appear in one's CZF. Thanks, Jeff On Wed, Mar 29, 2017 at 9:12 AM, John Shen (weifensh) <weife...@cisco.com> wrote: Hi Jeff, Thank you for the detail. I am wondering why there are two sets of lat/lang, i.e. one in CZF, the other is in Ops GUI Cache Group setting. To calculate the closest CG when matched CG in CZF is not available, the source lat/long is from mathced CZF, and the dest lat/long is from Ops setting, which doesnt seem to be consistent. Is there any reason why TR has this behavior? Since there are two sets of lat/long in TR, can we just use the lat/long all from Ops CG settings to get the closest, and do not care about the values set in CZF? At least this will avoid inconsistent config for lat/long. Thanks, John ---Original--- From: "Jeff Elsloo "<els...@apache.org> Date: 2017/3/29 22:45:12 To: "dev@trafficcontrol.incubator.apache.org"<dev@trafficcontrol .incubator. apache.org>; Subject: Re: Backup Cache Group Selection Yes, it's expected behavior. What you're describing sounds like a cachegroup in the CZF without any corresponding configuration in Traffic Ops, or a cachegroup with configuration in Traffic Ops, but with no available caches (DS assignments, health, etc). Presuming we have configured geolocation coordinates within the CZF, we know the lat/long of the cachegroup within the CZF that contains the source address. We can then order our list of cachegroups by lat/long, then select the "next best" cache group by distance and availability. That will be the actual cachegroup to serve the request; this prevents a miss on the CZF that would normally be routed to the Geolocation service selected for the DS. We do have a slight gap around logging, and maybe that's part of the question. What we see in the log is the selected lat/long, not the source lat/long of the hit, so we can't easily tell when we're in this case by simply looking at logs. This could be an area of improvement, however, we'll need to be careful to not conflate the logs with unnecessary information. In most cases the hit is the selected cachegroup, so we need to be careful to not just add "source" and "actual" coordinates to the log because it'll be identical in most CZF hit cases. Thanks, Thanks, Jeff On Wed, Mar 29, 2017 at 7:02 AM, John Shen (weifensh) <weife...@cisco.com> wrote: Hi Jeff, I have just tried the getClosestCacheLocation() logic. It appears the CZF matched lat/long does come from CZF, but the lat/long of the “closest” Cache Groups is from the configuration by Ops. This means to calculate the distance from the matched CG and “closest” CG, the source lat/long is from CZF, but the dest lat/long is not from CZF but from CG settings on Ops. Is this expected behavior? Thanks, John On 27/01/2017, 10:51 PM, "Jeff Elsloo" <jeff.els...@gmail.com> wrote: Steve: I don't think the patch is required, however, as Eric found, without the patch there could be some gaps depending on the scenario. That specific scenario revolved around the "next best cache group" not having a DS assigned, or a healthy cache with the DS assigned. In that case, despite the hits, you would still end up falling through to the geolocation provider. The patch addresses that. Eric: The rloc field is set via the Geolocation associated with the CacheLocation, which ultimately comes from the edgeLocations section of the CRConfig. When a CZF lookup is performed inside TR, a hit returns a CacheLocation. When caches aren't available within that CacheLocation, getClosestCacheLocation() is called, and that's why you see the lat/long of the "next best cache group" instead of the actual hit's lat/long. If we want to have granularity in this situation, we might need to 1) create a new RestultType, such as ResultType.CZ_NEXT (or something), and/or 2) massage the log format such that we either have a the original lat/long, and new lat/long in the rloc field, or create a new field to save one or the other, such that we log both lat/longs. Thoughts? Whatever we decide should go into TC-90 so we can apply the proposed patch and improve the logging. -- Thanks, Jeff On Fri, Jan 27, 2017 at 7:14 AM, Eric Friedrich (efriedri) <efrie...@cisco.com> wrote: The rloc field usually indicates the Geolocation IP of the client (short for request location) But here it looks like rloc is reflecting the location of the CG it ultimately redirected to (response location?). I would have expected the rloc field to either 1) be blank (because we never did a lookup from geoprovider) or 2) to contain the coordinates of the cache group the CZF hit on (in this case us-ga-macon at 32.7261, -83.6547”) —Eric On Jan 27, 2017, at 8:28 AM, Steve Malenfant < smalenf...@gmail.com> wrote: Jeff, CZF properly installed: yes Network address or not: same behavior But you nailed the API one. There is no cache assigned to us-ga-macon, which is exactly what I'm testing. I added cache groups for my testing in the lab which I assigned a few caches to them : - us-ga-atlanta 34.0362 -84.3207 - us-ok-oklahomacity 35.4777 -97.5545 - us-va-nova 38.7922 -77.2136 - us-ca-sandiego 32.7205 -117.0838 API : {"locationByGeo":{"city":"Macon","countryCode":"US"," latitude":"32.7288","postalCode":"31216","countryName":"United States","longitude":"-83.6865"},"locationByFederation":"not found","requestIp":"24.252.192.1","locationByCoverageZone":"not found"} Using the X-MM-Client-IP it returned the proper cache based on CZ, it correctly sent the request to the cache in us-ga-atlanta : 1485522786.423 qtype=HTTP chi=24.252.192.1 url=" http://crs.cox-col-jitp2.cdn1.coxlab.net/" cqhm=GET cqhv=HTTP/1.1 rtype=CZ rloc="34.03,-84.32" rdtl=- rerr="-" rgb="-" pssc=302 ttms=0.260 rurl=" http://cdn1cdedge0007.cox-col-jitp2.cdn1.coxlab.net/" rh="-" I then changed the coordinate to match the us-ca-sandiego group in the CZF and now the request is sent to the us-ca-sandiego caches : 1485523546.345 qtype=HTTP chi=24.252.192.1 url=" http://crs.cox-col-jitp2.cdn1.coxlab.net/" cqhm=GET cqhv=HTTP/1.1 rtype=CZ rloc="32.72,-117.08" rdtl=- rerr="-" rgb="-" pssc=302 ttms=0.206 rurl=" http://cdn1cdedge0001.cox-col-jitp2.cdn1.coxlab.net/" rh="- I'm using 1.6.1 + patch discussed in this email. Not sure if those are necessary but I'll need to try on unpatched version. Do we want to fix API to reflect CZF? Thanks for your help. Steve On Thu, Jan 26, 2017 at 4:47 PM, Jeff Elsloo <jeff.els...@gmail.com> wrote: Dave just let me know that in this case you don't have any caches assigned in us-ga-macon. I'm not sure how the API behaves at that point – it likely won't follow the same "next best cache group" logic, as it was designed as a simple lookup tool. Can you try simulating a request through Traffic Router directly using the X-MM-Client-IP header, or fakeClientIpAddress query parameter using the example IP of 24.252.192.0? After you do so, check the coordinates in the log entry and see if the result is a CZ hit. -- Thanks, Jeff On Thu, Jan 26, 2017 at 2:03 PM, Jeff Elsloo <jeff.els...@gmail.com> wrote: Are you 100% sure that the Traffic Router has loaded the updated CZF? If so, what happens when you use an IP within the /20 instead of the network address (.0)? I tried using a network address of a /22 on a 1.8 TR and it hit the CZF as expected. Ultimately what you're seeing is a CZF miss, unrelated to the geo coordinates. The underlying feature with the coordinates is to select the next best cache group by proximity where healthy caches have a given delivery service assigned. In order to test that, you would need to have a CZF hit in a cache group which doesn't have that particular delivery service assigned to any caches, or have all caches within that cache group with that delivery service in an unhealthy state. Thanks, -- Thanks, Jeff On Wed, Jan 25, 2017 at 1:33 PM, Steve Malenfant <smalenf...@gmail.com> wrote: Jeff, I've tried this coverage zone file coordinate overwrite... I might be missing something. I defined the following : "us-ga-macon": { "coordinates": { "latitude": "32.7261", "longitude": "-83.6547" }, "network": [ "24.252.192.0/20", "68.1.20.0/22", Then issued the following query : curl http://traffic_router:3333/crs/stats/ip/24.252.192.0 {"locationByGeo":{"city":"Macon","countryCode":"US"," latitude":"32.7288","postalCode":"31216","countryName":"United States","longitude":"-83.6865"},"locationByFederation":"not found","requestIp":"24.252.192.0"," locationByCoverageZone":"not found"} I believe I'm expecting "locationByCoverageZone" to find something... I tried on 1.6.0 and 1.6.1 (patched with the pastebin above which I wasn't sure I was suppose to do). Would you mind giving me some light on this? Thanks, Steve On Mon, Jan 23, 2017 at 3:05 PM, Jeff Elsloo <jeff.els...@gmail.com> wrote: Yes; the feature went into 1.5.x. -- Thanks, Jeff On Thu, Jan 19, 2017 at 10:37 AM, Steve Malenfant < smalenf...@gmail.com> wrote: I didn't know about this which is good information. Does that work on Traffic Router 1.6? On Mon, Jan 9, 2017 at 12:44 PM, Eric Friedrich (efriedri) < efrie...@cisco.com> wrote: Jeff and I had a quick Slack convo, so I’ll add a followup summary here in case anyone else is interested. Cache Group location (lat/long) is configured in Traffic Ops today (and is used for computing distance from Maxmind Geolocation). You can also configure the location (lat/long) for a Cache Group in the CoverageZone file (example below). When this location is configured (and Jeff’s suggested logic fix from below is applied) and all caches in the mapped cache group are unavailable, TR will send a client request to the cache group that is closest to the original mapped group. Example CZF w/ cache location ----- "coverageZones": { “edge-cg-1": { "network6": [ ... ], "network": [ ... ], "coordinates": { "longitude": “-75.3342", "latitude": “42.555" } }, —Eric On Jan 5, 2017, at 12:06 PM, Jeff Elsloo <jeff.els...@gmail.com> wrote: If we applied the proposed change, given your scenario we should fall through to the return statement that calls getClosestCacheLocation(). That method will order all cache groups based on their lat/long and the lat/long of the cache group we hit on in the CZF. Once the list is ordered, we iterate through the list until we find a cache group that has available caches for that DS. BTW, the stuff on line 536 is likely to produce the exact same result as the check that precedes it. networkNode.getLoc() will return the string name of the cache group, so when we find the CacheLocation, it will be the same as what we had just checked. We could probably get away with removing that part of the method as it's redundant. -- Thanks, Jeff On Wed, Jan 4, 2017 at 11:54 AM, Eric Friedrich (efriedri) <efrie...@cisco.com> wrote: Where would TR look outside the assigned cache group to find the next closest cache group? On Jan 4, 2017, at 11:25 AM, Eric Friedrich (efriedri) < efrie...@cisco.com> wrote: On Jan 3, 2017, at 5:20 PM, Jeff Elsloo <jeff.els...@gmail.com <mailto: jeff.els...@gmail.com>> wrote: Hey Eric, It sounds like the use case you're after is an RFC 1918 client associated with a cache group whose caches are all unavailable for one reason or another. Is that correct? Yes, exactly. I looked at the code a bit, and I think that we can make a minor change to achieve the behavior you're looking for as long as you're able to put your RFC 1918 ranges in the CZF. Yes, we would want those ranges in the CZF. I can’t think of any other place they would go. There's a small logic gap in the existing algorithm around cache location selection and I think if we fix that (two line change), we should be better off all around. I think the only time we'd ever want to go to the geolocation provider is in the event of a miss on the CZF, so as long as we have a hit there, we should find the cache group closest to that hit location that has available caches. This would automatically provide the "backup" cache group concept, and has the added benefit of doing this selection dynamically based on the state of the CDN. Wow, thanks for picking up on this solution. Sounds like a strong possibility. I like that it can extend dynamically. See this to get an idea of what I mean: http://apaste.info/u3PQo https://github.com/apache/ incubator-trafficcontrol/blob/ 249bd7504eeb7cc43402126f3719017e2475ad33/traffic_router/ core/src/main/java/com/comcast/cdn/traffic_control/ traffic_router/core/router/TrafficRouter.java#L536 Does this line set cacheLocation to the closest cache group with active caches on that DS? What does networkNode.getLoc() actually return? —Eric Obviously we'd need to test this to ensure we don't break other functionality. -- Thanks, Jeff On Tue, Jan 3, 2017 at 10:07 AM, Eric Friedrich (efriedri) <efrie...@cisco.com<mailto:efrie...@cisco.com>> wrote: If all caches in the primary cache group are unavailable, our goal is to provide a backup routing policy for RFC1918 clients. When client IP is an public Internet IP, the current backup policy is to assign the client to the geographically closest cache (Distance = MaxMind Geo Lat/Long - configured CG lat/long). When client IP is an RFC1918 IP, the client would not have a maxmind geo-loc, so would fall back to the DS geo-miss lat long. We’d prefer some more granular control over where these clients are routed to, rather than a per-DS setting. So with an RFC1918 client, the lookup process would be (step 3 is only addition) 1) Check CZF for a subnet match (and find a match for existing cache group). Assign client to CG 2) Check CG for available (online and associated w/ DS) servers. In this particular case, assume CG has no servers available to route the client to 3) Walk the CZF's list of backup CGs and perform the check from #2 for each CG. Use first server that is found 4) Assuming no server is found in #3, perform geo-location and find closest cache group. Use a server from the closest CG if one is found 4a) If geo-location returns null, use the DS’ default geo-miss location as the client location. —Eric On Dec 26, 2016, at 10:01 AM, Jan van Doorn <j...@knutsel.com <mailto: j...@knutsel.com>> wrote: Hi Eric, How does the backup list relate to the RFC1918-is-not-in-geo problem? To get to a cachegroup you need to get a match in the coverage zone, I would think? Rgds, JvD On Dec 22, 2016, at 12:28, Eric Friedrich (efriedri) < efrie...@cisco.com<mailto:efrie...@cisco.com>> wrote: The current behavior of cache group selection works as follows 1) Look for a subnet match in CZF 2) Use MaxMind/Neustar for GeoLocation based on client IP. Choose closest cache group. 3) Use Delivery Service Geo-Miss Lat/Long. Choose closest cache group. For deployments where IP addressing is primarily private (say RFC-1918 addresses), client IP Geo Location (#2) is not useful. We are considering adding another field to the Coverage Zone File that configures an ordered list of backup cache groups to try if the primary cache group does not have any available caches. Example: "coverageZones": { "cache-group-01": { “backupList”: [“cache-group-02”, “cache-group-03”], "network6": [ "1234:5678::\/64”, "1234:5679::\/64"], "network": [ "192.168.8.0\/24", "192.168.9.0\/24”] } This configuration could also be part of the per-cache group configuration, but that would give less control over which clients preferred which cache groups. For example, you may have cache groups in LA, Chicago and NY. If the Chicago Cache group fails, you may want some of the Chicago clients to go to LA and some to go to NY. If the backup CG configuration is per-cg, we would not be able to control where clients are allocated. Looking for opinions and comments on the above proposal, this is still in idea stage. Thanks All! Eric -- *Oren Shemesh* Qwilt | Work: +972-72-2221637 <072-222-1637>| Mobile: +972-50-2281168 <050-228-1168> | or...@qwilt.com<mailto:or...@qwilt.com> <y...@qwilt.com<mailto:y...@qwilt.com>> -- *Ori Finkelman*Qwilt | Work: +972-72-2221647 <072-222-1647> | Mobile: +972-52-3832189 <052-383-2189> | o...@qwilt.com<mailto:o...@qwilt.com> -- *Ori Finkelman*Qwilt | Work: +972-72-2221647 | Mobile: +972-52-3832189 | o...@qwilt.com<mailto:o...@qwilt.com>